Vibecode Chatbase

track this build5 phases, 16 steps, beginner friendly

KINDA · weekend project

price $150/moyou'd save $1,800/yrbuild time multi-day for one web agent; multi-week for the platformcategory 🛟 customer supportreplaced by 0 people

A useful web version is a serious multi-day build, not magic: crawl or upload sources, retrieve tenant-scoped passages, stream cited answers, run one guarded business action, and hand uncertain requests to a person. Rebuilding Chatbase itself is a different project. Its paid surface combines source management and retraining, procedures and custom UI actions, a polished widget, helpdesk, identity, analytics, voice, email, social and CRM channels, plus the reliability and security work that keeps an agent safe in front of customers.

the project pack5 files · written for this build, step by step

You are building a lean indie version of Chatbase.

Create the following project files first, then implement the application by following them. Keep the files updated as decisions change. Do not collapse this into a single README or prompt.

===== README.md =====
# Chatbase · indie build

Build a bounded support FAQ assistant with citations and a human handoff queue.

Estimated effort: **multi-day**. Work `BUILD_PLAN.md` top to bottom · every phase ends in a check that has to pass before the next one starts.

## Stack

| Part | Choice | Why |
| --- | --- | --- |
| Runtime | Python 3.12+ and Flask | A small local HTTP application with readable server-side handlers and no JavaScript build pipeline. |
| Storage | SQLite through Python sqlite3 | Local ownership with explicit transactions and a portable backup path. |
| Interface | Jinja HTML templates, CSS and vanilla JavaScript | Implement the specified screens directly with labelled controls and progressive enhancement. |
| Verification | Python unittest and browser acceptance fixtures | Check invariants and the actual user workflow before the next phase. |
| Generation | Ollama local HTTP API | A backend-only adapter with a configurable installed local model and visible failure states. |

## Before you start

Have every one of these ready. The plan assumes them from step one.

- [ ] **Python 3.12 or newer and a terminal** · Python is free; this local prototype requires no hosting subscription.
- Why: The local backend, tests and storage use Python.
- Get it: Install Python from python.org. Open a terminal in a new empty project folder. On Windows, use py -3 where this guide says python if python is not on PATH; on macOS/Linux use python3 if needed. Create .venv and activate it before installing packages.
- Verify: python --version reports 3.12+ and python -c "import sqlite3" succeeds.
- [ ] **A current desktop browser and text editor** · No paid service required for this local step.
- Why: You need to view the local interface and edit generated project files.
- Get it: Use an installed Chrome, Firefox or Edge browser. Open the empty project folder in your editor; open the terminal there, not in this catalog repository.
- Verify: You can create a text file in the project folder and open browser developer tools.
- [ ] **Chatbase acceptance sample** · No paid service required for this local step.
- Why: This fixture makes the first end-to-end check reproducible.
- Get it: Create synthetic sample data before implementation: Five supplied FAQ documents, an answerable return-policy question, an unknown question and a source containing hostile instructions. Save it under fixtures/ with the expected results in fixtures/README.md; use no real credentials or customer records.
- Verify: The fixture files and expected outcomes exist and can be inspected without a vendor account.
- [ ] **Confirm the bounded replacement** · No paid service required for this local step.
- Why: The full commercial product is larger than this guide.
- Get it: The deliverable is a bounded support FAQ assistant with citations and a human handoff queue. Read every exclusion below. Indie mode runs locally; any public or shared use requires the product access and deployment gates first.
- Verify: You can name the primary workflow and the vendor features this version deliberately leaves out.
- [ ] **Ollama with an explicitly local model** · No cloud inference fee for this local configuration; you supply disk space, compute and electricity.
- Why: Generation requires a running model; ordinary editing remains usable while it is unavailable.
- Get it: Install Ollama from the official site, choose a locally runnable model from its library that fits available RAM, download it, and copy its exact installed name into LOCAL_MODEL. Use the documented local-only configuration, exclude cloud model names, and record model license, download size and tested memory use.
- Verify: ollama list shows the chosen model; a short local prompt succeeds with network access disabled after download.

## Quick start

```sh
python -m venv .venv
python -m pip install Flask Flask-WTF python-dotenv waitress tzdata
python -m pip freeze > requirements.txt
python -m flask --app app run --host 127.0.0.1
```

Then copy `.env.example` to `.env` and fill in the values it documents.

## Honest limits

This build deliberately does not replace:

- Voice, write actions, customer account tools and omnichannel support are excluded; this FAQ agent cannot change orders.
- This is a bounded support FAQ assistant with citations and a human handoff queue; full Chatbase feature parity, vendor hosting and support are not included.
- the no-code agent lifecycle: many managed source connectors, background retraining, source suggestions, and production ingestion diagnostics
- the channel network: website chat, email, voice and telephony, Slack, WhatsApp, Messenger, Instagram, Shopify, WordPress, Zapier, Zendesk, Salesforce, and other integrations
- the support operation around the agent: a team helpdesk, tickets, assignment, contact identity, authenticated personalization, escalation, and access controls
- the action and procedure builder: server, client, button, and custom widget actions with testing, permissions, and integration-specific behavior
- the production layer: topic and sentiment analytics, review workflows, abuse controls, delivery retries, observability, data governance, compliance, and support

If one of those is essential to you, that is the reason to keep paying for Chatbase, and the README should say so rather than pretend.

===== BRIEF.md =====
# Build brief · Chatbase

The one-shot brief this plan expands. `BUILD_PLAN.md` (or `MILESTONES.md`) is the same sequence broken into steps and checks; where the two disagree, the plan wins.

Build me a bounded support FAQ assistant with citations and a human handoff queue as a bounded replacement for Chatbase.
Build it in phases in the order below. Do not write the whole product in one pass.
Finish the current phase, run its Done when checks, fix failures, then continue.
Indie mode completes Phases 1-5. Phase 6 is required before shared or public deployment.

### Stack (fixed, do not substitute)
- Runtime: Python 3.12+ and Flask. A small local HTTP application with readable server-side handlers and no JavaScript build pipeline.
- Storage: SQLite through Python sqlite3. Local ownership with explicit transactions and a portable backup path.
- Interface: Jinja HTML templates, CSS and vanilla JavaScript. Implement the specified screens directly with labelled controls and progressive enhancement.
- Verification: Python unittest and browser acceptance fixtures. Check invariants and the actual user workflow before the next phase.
- Generation: Ollama local HTTP API. A backend-only adapter with a configurable installed local model and visible failure states.

### Data model (create this before Phase 1)
- sources(id,title,url,text,hash,enabled)
- chunks(id,source_id,text,position)
- conversations(id,visitor_hash)
- messages(id,conversation_id,role,text)
- citations(message_id,chunk_id)
- handoffs(id,conversation_id,email,summary,state)
- Use stable ids and explicit relationships; add auxiliary tables named by the phases before those phases use them.
- Acceptance fixture: Five supplied FAQ documents, an answerable return-policy question, an unknown question and a source containing hostile instructions.

### Prerequisites and configuration
- Python 3.12 or newer and a terminal: Install Python from python.org. Open a terminal in a new empty project folder. On Windows, use py -3 where this guide says python if python is not on PATH; on macOS/Linux use python3 if needed. Create .venv and activate it before installing packages. Verify: python --version reports 3.12+ and python -c "import sqlite3" succeeds. Cost: Python is free; this local prototype requires no hosting subscription.
- A current desktop browser and text editor: Use an installed Chrome, Firefox or Edge browser. Open the empty project folder in your editor; open the terminal there, not in this catalog repository. Verify: You can create a text file in the project folder and open browser developer tools. Cost: No paid service required for this local step.
- Chatbase acceptance sample: Create synthetic sample data before implementation: Five supplied FAQ documents, an answerable return-policy question, an unknown question and a source containing hostile instructions. Save it under fixtures/ with the expected results in fixtures/README.md; use no real credentials or customer records. Verify: The fixture files and expected outcomes exist and can be inspected without a vendor account. Cost: No paid service required for this local step.
- Confirm the bounded replacement: The deliverable is a bounded support FAQ assistant with citations and a human handoff queue. Read every exclusion below. Indie mode runs locally; any public or shared use requires the product access and deployment gates first. Verify: You can name the primary workflow and the vendor features this version deliberately leaves out. Cost: No paid service required for this local step.
- Ollama with an explicitly local model: Install Ollama from the official site, choose a locally runnable model from its library that fits available RAM, download it, and copy its exact installed name into LOCAL_MODEL. Use the documented local-only configuration, exclude cloud model names, and record model license, download size and tested memory use. Verify: ollama list shows the chosen model; a short local prompt succeeds with network access disabled after download. Cost: No cloud inference fee for this local configuration; you supply disk space, compute and electricity.
- Configuration: DATA_DIR, SECRET_KEY, APP_MODE, SITE_URL, OLLAMA_URL, LOCAL_MODEL. Ship .env.example, replace placeholders and exclude secrets from git.

### Phase 1 · Local foundation and data contract
Build: Create the runnable local foundation for a bounded support FAQ assistant with citations and a human handoff queue, including the real schema and acceptance fixture.

- Run the venv command first, then stop and activate it before running the remaining commands. Use your OS instructions: Windows .venv\Scripts\activate.bat in cmd, or .venv\Scripts\Activate.ps1 in PowerShell; macOS/Linux source .venv/bin/activate. Then install the listed packages. Create app.py exposing app, load .env with python-dotenv, and add a /healthz route. Freeze tested dependency versions to requirements.txt. Create .env manually from .env.example; replace SECRET_KEY before launch. Commands, after creating the named files: `python -m venv .venv`; `python -m pip install Flask Flask-WTF python-dotenv waitress tzdata`; `python -m pip freeze > requirements.txt`; `python -m flask --app app run --host 127.0.0.1`.
- For database-backed state create schema.sql and db.py using this model: sources(id,title,url,text,hash,enabled); chunks(id,source_id,text,position); conversations(id,visitor_hash); messages(id,conversation_id,role,text); citations(message_id,chunk_id); handoffs(id,conversation_id,email,summary,state). Add the auxiliary relationships explicitly used in later phases; use stable ids, foreign keys, parameterized values and transactions. For canonical files use atomic replacement and hashes instead. Implement python manage.py init and python manage.py seed to create only synthetic data: Five supplied FAQ documents, an answerable return-policy question, an unknown question and a source containing hostile instructions. Re-running init must not erase data. Commands, after creating the named files: `python manage.py init`; `python manage.py seed`.
- Install Flask-WTF CSRF protection for forms and same-origin fetches, validate Host and Origin against SITE_URL, bound inputs and escape output. Deliberately public intake endpoints get bounded validation and rate limits; signed provider webhooks use signature checks instead of browser CSRF tokens. All other admin routes stay local until product authentication exists. When a later phase tests viewer/member permissions, implement the users, hashed passwords, expiring server sessions and role checks in this foundation first; product mode must never use a fabricated identity from request input. Add error, empty, keyboard-focus and retry states now.
- Create providers/local_model.py with a POST /api/chat adapter using the configured local model and stream:false. Validate structured output, bound input/output, set a timeout and reject cloud models or non-loopback inference URLs. Add a probe that sends one harmless prompt and a test double for deterministic tests. Missing model setup must disable generation with a specific error while leaving editing usable. Commands, after creating the named files: `python scripts/model_probe.py`.
- Trap: The Flask development server is for local development only. Loopback binding alone does not replace CSRF and origin validation.

Done when: The documented start command serves /healthz and the initial screen at http://127.0.0.1:5000. Restarting preserves the seeded records or canonical files; running init again deletes nothing. A cross-origin state-changing request to an admin route is refused.
Do not build yet: Do not start controlled knowledge intake until these checks pass.

### Phase 2 · Controlled knowledge intake
Build: Controlled knowledge intake for Chatbase: Accept owner-uploaded TXT/Markdown and pasted text with provenance; split into bounded paragraphs and index locally.

- Accept owner-uploaded TXT/Markdown and pasted text with provenance; split into bounded paragraphs and index locally.
- Preview source content and allow enable/disable/delete; delete associated chunks when removing a source.
- Use Five supplied FAQ documents, an answerable return-policy question, an unknown question and a source containing hostile instructions. Check: Reimporting an unchanged source does not duplicate chunks. Then check: Disabling a source removes it from subsequent retrieval. Add a focused regression for the failure case under test/ using a temporary database or folder and fake external services; also perform the visible workflow in a browser. Commands, after creating the named files: `python -m unittest discover -s test`.
- Trap: No open-web crawler, document execution or automatic connector is included in this first version.

Done when: Reimporting an unchanged source does not duplicate chunks. Disabling a source removes it from subsequent retrieval.
Do not build yet: Do not start cited answers until these checks pass.

### Phase 3 · Cited answers
Build: Cited answers for Chatbase: Retrieve matching passages with SQLite full-text search and supply only those passages to the local model.

- Retrieve matching passages with SQLite full-text search and supply only those passages to the local model.
- Require source chunk ids in the answer, reject unknown ids and show an abstention when evidence is absent; render output as escaped text.
- Use Five supplied FAQ documents, an answerable return-policy question, an unknown question and a source containing hostile instructions. Check: The return-policy answer cites an existing enabled chunk containing the policy. Then check: The unknown question produces an abstention, and hostile source instructions cannot trigger tools. Add a focused regression for the failure case under test/ using a temporary database or folder and fake external services; also perform the visible workflow in a browser. Commands, after creating the named files: `python -m unittest discover -s test`.
- Trap: A valid citation id alone does not prove entailment; keep an owner-reviewed answer evaluation set.

Done when: The return-policy answer cites an existing enabled chunk containing the policy. The unknown question produces an abstention, and hostile source instructions cannot trigger tools.
Do not build yet: Do not start widget and handoff until these checks pass.

### Phase 4 · Widget and handoff
Build: Widget and handoff for Chatbase: Build an iframe chat widget with bounded messages, signed conversation identity and explicit allowed embedding origins.

- Build an iframe chat widget with bounded messages, signed conversation identity and explicit allowed embedding origins.
- Offer a consented email/summary handoff form stored in an owner queue; provide conversation export and retention deletion.
- Use Five supplied FAQ documents, an answerable return-policy question, an unknown question and a source containing hostile instructions. Check: One signed visitor cannot read another conversation by changing its id. Then check: Requesting a person creates one handoff and does not invent an order lookup result. Add a focused regression for the failure case under test/ using a temporary database or folder and fake external services; also perform the visible workflow in a browser. Commands, after creating the named files: `python -m unittest discover -s test`.
- Trap: Voice, write actions, customer account tools and omnichannel support are excluded; this FAQ agent cannot change orders.

Done when: One signed visitor cannot read another conversation by changing its id. Requesting a person creates one handoff and does not invent an order lookup result.
Do not build yet: Do not start portable delivery and clean-start proof until these checks pass.

### Phase 5 · Portable delivery and clean-start proof
Build: Finish the Chatbase replacement with a documented first-run path and a recoverable export.

- Document prerequisites, every configuration variable, data locations and the three domain workflows: Controlled knowledge intake; Cited answers; Widget and handoff. Include the exclusions, actual costs and source check date 2026-09-05. Commands shown earlier that reference project files are files the builder must implement before running them.
- Use sqlite3.Connection.backup for the database in DATA_DIR and pause writers while taking the matching attachment snapshot. Keep schema version, hashes and a manifest for chatbase. Restore into a separate empty directory or isolated deployment, with all outbound actions and schedules disabled. Recreate this proof: Five supplied FAQ documents, an answerable return-policy question, an unknown question and a source containing hostile instructions. Compare ids/counts and open the result before recording the restore date. Never test a restore over the only live copy.
- Use a new folder and only README instructions to reach the fixture outcome. Install from requirements.txt, initialize storage, run the tests and exercise the UI at 375px and 1280px; use keyboard-only controls and inspect browser errors. Record exactly what passed; do not mark unrun checks complete. Commands, after creating the named files: `python -m pip install -r requirements.txt`; `python -m unittest discover -s test`.
- Trap: Export without an import or restore rehearsal is not verified portability.

Done when: A new folder reaches the documented fixture outcome using only README instructions. A backup restored in isolation preserves the proof data and can be opened.
Do not build yet: Do not start operate and share with access controls until these checks pass.

### Phase 6 · Operate and share with access controls (product-builder only)
Build: Product-builder only: make a bounded support FAQ assistant with citations and a human handoff queue operable for its stated audience before allowing remote access.

- Add owner/member/viewer authorization as appropriate, password hashes through Werkzeug, expiring server-side sessions, CSRF and login rate limits. Add users, sessions and audit schema migrations plus an owner-init command prompting securely for its password. Enforce ownership on every private read/write/download and fail closed when product setup is incomplete. Existing learner or admin identities must use this same boundary.
- Serve app:app with Waitress on loopback behind Caddy HTTPS. Configure trusted proxy headers narrowly, Secure/HttpOnly/SameSite cookies, upload limits, request ids and structured logs without private content. Run background jobs as a separate supervised process when the domain workflow requires them. Commands, after creating the named files: `waitress-serve --listen=127.0.0.1:5000 app:app`.
- Add a scheduled encrypted off-host backup using this procedure: Use sqlite3.Connection.backup for the database in DATA_DIR and pause writers while taking the matching attachment snapshot. Keep schema version, hashes and a manifest for chatbase. Monitor health, disk space, failed jobs and backup age. Stop the service to test an alert; rehearse recovery with outbound effects disabled and record recovery time. Rotate any exposed credentials and require reconciliation before resuming queues.
- Trap: Product mode is additional engineering, not a switch that makes a local prototype production-ready.

Done when: An unauthenticated visitor cannot read private records, download private files or mutate the admin surface. A viewer cannot mutate data by making a direct request, and guessed ids cannot cross ownership boundaries. Stopping the service triggers the configured alert and the latest off-host backup restores the fixture in isolation.
Do not build yet: Do not add new providers or vendor-scale features until the release gates are recorded.

### Out of scope (and why)
- Voice, write actions, customer account tools and omnichannel support are excluded; this FAQ agent cannot change orders.
- This is a bounded support FAQ assistant with citations and a human handoff queue; full Chatbase feature parity, vendor hosting and support are not included.
- the no-code agent lifecycle: many managed source connectors, background retraining, source suggestions, and production ingestion diagnostics: retain the paid product if this capability is essential; it is outside this bounded build.
- the channel network: website chat, email, voice and telephony, Slack, WhatsApp, Messenger, Instagram, Shopify, WordPress, Zapier, Zendesk, Salesforce, and other integrations: retain the paid product if this capability is essential; it is outside this bounded build.
- the support operation around the agent: a team helpdesk, tickets, assignment, contact identity, authenticated personalization, escalation, and access controls: retain the paid product if this capability is essential; it is outside this bounded build.
- the action and procedure builder: server, client, button, and custom widget actions with testing, permissions, and integration-specific behavior: retain the paid product if this capability is essential; it is outside this bounded build.
- the production layer: topic and sentiment analytics, review workflows, abuse controls, delivery retries, observability, data governance, compliance, and support: retain the paid product if this capability is essential; it is outside this bounded build.

### README must contain
- Exact install, configure, initialize, start, test and recovery commands for a clean folder.
- The schema, fixture, acceptance results, private data locations and export format.
- Every prerequisite and secret source, actual running costs and deliberately excluded capabilities.
- An honest distinction between tests actually run and checks still awaiting real credentials or deployment.
- Dependency documentation reviewed 2026-09-05: https://flask.palletsprojects.com/en/stable/installation/ ; https://docs.python.org/3/library/sqlite3.html ; https://flask-wtf.readthedocs.io/en/1.2.x/csrf/ ; https://docs.pylonsproject.org/projects/waitress/en/stable/usage.html ; https://docs.ollama.com/api/chat ; https://docs.ollama.com/cloud. Recheck before upgrading; freeze the versions actually tested.

===== AGENTS.md =====
# Agent instructions · Chatbase indie build

- Read `README.md` and `BUILD_PLAN.md` before writing code. The stack is fixed: Python 3.12+ and Flask, SQLite through Python sqlite3, Jinja HTML templates, CSS and vanilla JavaScript, Python unittest and browser acceptance fixtures, Ollama local HTTP API. Do not substitute.
- Work one phase at a time, in order. Do not start a phase until every "Done when" item of the previous one passes.
- Prefer the fewest moving parts that satisfy the step. No frameworks, services or dependencies the plan does not name.
- Secrets live in `.env`, never in source or logs. Keep `.env.example` current when a variable is introduced.
- Do not invent cryptography, security guarantees, APIs or compliance claims.
- Add a focused test for every destructive, security-sensitive or data-loss path the plan names.
- Run the project checks before declaring a phase complete, and record any deliberate shortcut in the README under "Tradeoffs".

## Known traps

- The Flask development server is for local development only. Loopback binding alone does not replace CSRF and origin validation.
- No open-web crawler, document execution or automatic connector is included in this first version.
- A valid citation id alone does not prove entailment; keep an owner-reviewed answer evaluation set.
- Voice, write actions, customer account tools and omnichannel support are excluded; this FAQ agent cannot change orders.
- Export without an import or restore rehearsal is not verified portability.

===== BUILD_PLAN.md =====
# Build plan · Chatbase

Build a bounded support FAQ assistant with citations and a human handoff queue.

Phases are in dependency order. Each ends in a "Done when" list; treat an unticked item as a blocker, not a note.

## Phase 1 · Local foundation and data contract

Create the runnable local foundation for a bounded support FAQ assistant with citations and a human handoff queue, including the real schema and acceptance fixture.

### Steps

1. Create and start the local application
Run the venv command first, then stop and activate it before running the remaining commands. Use your OS instructions: Windows .venv\Scripts\activate.bat in cmd, or .venv\Scripts\Activate.ps1 in PowerShell; macOS/Linux source .venv/bin/activate. Then install the listed packages. Create app.py exposing app, load .env with python-dotenv, and add a /healthz route. Freeze tested dependency versions to requirements.txt. Create .env manually from .env.example; replace SECRET_KEY before launch.
Files: `app.py`, `requirements.txt`, `.env.example`, `.gitignore`

```sh
python -m venv .venv
python -m pip install Flask Flask-WTF python-dotenv waitress tzdata
python -m pip freeze > requirements.txt
python -m flask --app app run --host 127.0.0.1
```

2. Implement storage and seed the named fixture
For database-backed state create schema.sql and db.py using this model: sources(id,title,url,text,hash,enabled); chunks(id,source_id,text,position); conversations(id,visitor_hash); messages(id,conversation_id,role,text); citations(message_id,chunk_id); handoffs(id,conversation_id,email,summary,state). Add the auxiliary relationships explicitly used in later phases; use stable ids, foreign keys, parameterized values and transactions. For canonical files use atomic replacement and hashes instead. Implement python manage.py init and python manage.py seed to create only synthetic data: Five supplied FAQ documents, an answerable return-policy question, an unknown question and a source containing hostile instructions. Re-running init must not erase data.
Files: `schema.sql`, `db.py`, `manage.py`, `fixtures/README.md`

```sh
python manage.py init
python manage.py seed
```

3. Protect the local request boundary
Install Flask-WTF CSRF protection for forms and same-origin fetches, validate Host and Origin against SITE_URL, bound inputs and escape output. Deliberately public intake endpoints get bounded validation and rate limits; signed provider webhooks use signature checks instead of browser CSRF tokens. All other admin routes stay local until product authentication exists. When a later phase tests viewer/member permissions, implement the users, hashed passwords, expiring server sessions and role checks in this foundation first; product mode must never use a fabricated identity from request input. Add error, empty, keyboard-focus and retry states now.
Files: `app.py`, `templates/base.html`, `test/test_foundation.py`
4. Verify the selected local model
Create providers/local_model.py with a POST /api/chat adapter using the configured local model and stream:false. Validate structured output, bound input/output, set a timeout and reject cloud models or non-loopback inference URLs. Add a probe that sends one harmless prompt and a test double for deterministic tests. Missing model setup must disable generation with a specific error while leaving editing usable.
Files: `providers/local_model.py`, `scripts/model_probe.py`

```sh
python scripts/model_probe.py
```

### Done when

- [ ] The documented start command serves /healthz and the initial screen at http://127.0.0.1:5000.
- [ ] Restarting preserves the seeded records or canonical files; running init again deletes nothing.
- [ ] A cross-origin state-changing request to an admin route is refused.

### Watch out

- The Flask development server is for local development only. Loopback binding alone does not replace CSRF and origin validation.

## Phase 2 · Controlled knowledge intake

Controlled knowledge intake for Chatbase: Accept owner-uploaded TXT/Markdown and pasted text with provenance; split into bounded paragraphs and index locally.

### Steps

1. Accept owner-uploaded TXT/Markdown and pasted text with provenance; split into bounded paragraphs and index locally.
Accept owner-uploaded TXT/Markdown and pasted text with provenance; split into bounded paragraphs and index locally.
Files: `features/controlled_knowledge_intake.py`, `templates/controlled_knowledge_intake.html`
2. Preview source content and allow enable/disable/delete; delete associated chunks when removing a source.
Preview source content and allow enable/disable/delete; delete associated chunks when removing a source.
Files: `features/controlled_knowledge_intake.py`
3. Exercise the acceptance fixture
Use Five supplied FAQ documents, an answerable return-policy question, an unknown question and a source containing hostile instructions. Check: Reimporting an unchanged source does not duplicate chunks. Then check: Disabling a source removes it from subsequent retrieval. Add a focused regression for the failure case under test/ using a temporary database or folder and fake external services; also perform the visible workflow in a browser.
Files: `test/test_controlled_knowledge_intake.py`

```sh
python -m unittest discover -s test
```

### Done when

- [ ] Reimporting an unchanged source does not duplicate chunks.
- [ ] Disabling a source removes it from subsequent retrieval.

### Watch out

- No open-web crawler, document execution or automatic connector is included in this first version.

## Phase 3 · Cited answers

Cited answers for Chatbase: Retrieve matching passages with SQLite full-text search and supply only those passages to the local model.

### Steps

1. Retrieve matching passages with SQLite full-text search and supply only those passages to the local model.
Retrieve matching passages with SQLite full-text search and supply only those passages to the local model.
Files: `features/cited_answers.py`, `templates/cited_answers.html`
2. Require source chunk ids in the answer, reject unknown ids and show an abstention when evidence is absent; render output as escaped text.
Require source chunk ids in the answer, reject unknown ids and show an abstention when evidence is absent; render output as escaped text.
Files: `features/cited_answers.py`
3. Exercise the acceptance fixture
Use Five supplied FAQ documents, an answerable return-policy question, an unknown question and a source containing hostile instructions. Check: The return-policy answer cites an existing enabled chunk containing the policy. Then check: The unknown question produces an abstention, and hostile source instructions cannot trigger tools. Add a focused regression for the failure case under test/ using a temporary database or folder and fake external services; also perform the visible workflow in a browser.
Files: `test/test_cited_answers.py`

```sh
python -m unittest discover -s test
```

### Done when

- [ ] The return-policy answer cites an existing enabled chunk containing the policy.
- [ ] The unknown question produces an abstention, and hostile source instructions cannot trigger tools.

### Watch out

- A valid citation id alone does not prove entailment; keep an owner-reviewed answer evaluation set.

## Phase 4 · Widget and handoff

Widget and handoff for Chatbase: Build an iframe chat widget with bounded messages, signed conversation identity and explicit allowed embedding origins.

### Steps

1. Build an iframe chat widget with bounded messages, signed conversation identity and explicit allowed embedding origins.
Build an iframe chat widget with bounded messages, signed conversation identity and explicit allowed embedding origins.
Files: `features/widget_and_handoff.py`, `templates/widget_and_handoff.html`
2. Offer a consented email/summary handoff form stored in an owner queue; provide conversation export and retention deletion.
Offer a consented email/summary handoff form stored in an owner queue; provide conversation export and retention deletion.
Files: `features/widget_and_handoff.py`
3. Exercise the acceptance fixture
Use Five supplied FAQ documents, an answerable return-policy question, an unknown question and a source containing hostile instructions. Check: One signed visitor cannot read another conversation by changing its id. Then check: Requesting a person creates one handoff and does not invent an order lookup result. Add a focused regression for the failure case under test/ using a temporary database or folder and fake external services; also perform the visible workflow in a browser.
Files: `test/test_widget_and_handoff.py`

```sh
python -m unittest discover -s test
```

### Done when

- [ ] One signed visitor cannot read another conversation by changing its id.
- [ ] Requesting a person creates one handoff and does not invent an order lookup result.

### Watch out

- Voice, write actions, customer account tools and omnichannel support are excluded; this FAQ agent cannot change orders.

## Phase 5 · Portable delivery and clean-start proof

Finish the Chatbase replacement with a documented first-run path and a recoverable export.

### Steps

1. Write the user-facing runbook
Document prerequisites, every configuration variable, data locations and the three domain workflows: Controlled knowledge intake; Cited answers; Widget and handoff. Include the exclusions, actual costs and source check date 2026-09-05. Commands shown earlier that reference project files are files the builder must implement before running them.
Files: `README.md`, `.env.example`
2. Create and rehearse a recoverable backup
Use sqlite3.Connection.backup for the database in DATA_DIR and pause writers while taking the matching attachment snapshot. Keep schema version, hashes and a manifest for chatbase. Restore into a separate empty directory or isolated deployment, with all outbound actions and schedules disabled. Recreate this proof: Five supplied FAQ documents, an answerable return-policy question, an unknown question and a source containing hostile instructions. Compare ids/counts and open the result before recording the restore date. Never test a restore over the only live copy.
Files: `OPERATIONS.md`, `scripts/backup.py`
3. Reproduce the first-run experience
Use a new folder and only README instructions to reach the fixture outcome. Install from requirements.txt, initialize storage, run the tests and exercise the UI at 375px and 1280px; use keyboard-only controls and inspect browser errors. Record exactly what passed; do not mark unrun checks complete.
Files: `README.md`, `verification.md`

```sh
python -m pip install -r requirements.txt
python -m unittest discover -s test
```

### Done when

- [ ] A new folder reaches the documented fixture outcome using only README instructions.
- [ ] A backup restored in isolation preserves the proof data and can be opened.

### Watch out

- Export without an import or restore rehearsal is not verified portability.

## Not in this build

## After v1, if you want it

- Extend widget and handoff only after the fixture and recovery checks pass.
- Before adding another integration, document its credentials, limits, cost and failure behavior with primary-source links.

===== .env.example =====
# Copy to .env and fill in. Never commit .env; this file documents it.

# Required. A new private folder this application creates on first run; never serve it as static files.
DATA_DIR=./data/chatbase

# Required · secret. Generate with python -c "import secrets; print(secrets.token_hex(32))" and save only in .env.
SECRET_KEY=replace-with-a-random-secret

# Required. Use indie for loopback-only personal work. Product mode must refuse to start until owner login and the deployment gates are configured.
APP_MODE=indie

# Required. The origin used for local forms and links; replace with the chosen HTTPS origin in product mode.
SITE_URL=http://127.0.0.1:5000

# Required. Keep the inference service bound to loopback; the backend calls it, never the public browser.
OLLAMA_URL=http://127.0.0.1:11434

# Required. Copy the exact non-cloud name from ollama list; preflight must reject this placeholder.
LOCAL_MODEL=replace-with-installed-local-model

You are building a lean indie version of Chatbase.

Create the following project files first, then implement the application by following them. Keep the files updated as decisions change. Do not collapse this into a single README or prompt.

===== README.md =====
# Chatbase · indie build

Build a bounded support FAQ assistant with citations and a human handoff queue.

Estimated effort: **multi-day**. Work `BUILD_PLAN.md` top to bottom · every phase ends in a check that has to pass before the next one starts.

## Stack

## Before you start

Have every one of these ready. The plan assumes them from step one.

## Quick start

```sh
python -m venv .venv
python -m pip install Flask Flask-WTF python-dotenv waitress tzdata
python -m pip freeze > requirements.txt
python -m flask --app app run --host 127.0.0.1
```

Then copy `.env.example` to `.env` and fill in the values it documents.

## Honest limits

This build deliberately does not replace:

If one of those is essential to you, that is the reason to keep paying for Chatbase, and the README should say so rather than pretend.

===== BRIEF.md =====
# Build brief · Chatbase

The one-shot brief this plan expands. `BUILD_PLAN.md` (or `MILESTONES.md`) is the same sequence broken into steps and checks; where the two disagree, the plan wins.

Done when: Reimporting an unchanged source does not duplicate chunks. Disabling a source removes it from subsequent retrieval.
Do not build yet: Do not start cited answers until these checks pass.

### Phase 3 · Cited answers
Build: Cited answers for Chatbase: Retrieve matching passages with SQLite full-text search and supply only those passages to the local model.

### Phase 4 · Widget and handoff
Build: Widget and handoff for Chatbase: Build an iframe chat widget with bounded messages, signed conversation identity and explicit allowed embedding origins.

### Phase 5 · Portable delivery and clean-start proof
Build: Finish the Chatbase replacement with a documented first-run path and a recoverable export.

===== AGENTS.md =====
# Agent instructions · Chatbase indie build

## Known traps

===== BUILD_PLAN.md =====
# Build plan · Chatbase

Build a bounded support FAQ assistant with citations and a human handoff queue.

Phases are in dependency order. Each ends in a "Done when" list; treat an unticked item as a blocker, not a note.

## Phase 1 · Local foundation and data contract

Create the runnable local foundation for a bounded support FAQ assistant with citations and a human handoff queue, including the real schema and acceptance fixture.

### Steps

```sh
python -m venv .venv
python -m pip install Flask Flask-WTF python-dotenv waitress tzdata
python -m pip freeze > requirements.txt
python -m flask --app app run --host 127.0.0.1
```

```sh
python manage.py init
python manage.py seed
```

```sh
python scripts/model_probe.py
```

### Done when

### Watch out

- The Flask development server is for local development only. Loopback binding alone does not replace CSRF and origin validation.

## Phase 2 · Controlled knowledge intake

Controlled knowledge intake for Chatbase: Accept owner-uploaded TXT/Markdown and pasted text with provenance; split into bounded paragraphs and index locally.

### Steps

```sh
python -m unittest discover -s test
```

### Done when

- [ ] Reimporting an unchanged source does not duplicate chunks.
- [ ] Disabling a source removes it from subsequent retrieval.

### Watch out

- No open-web crawler, document execution or automatic connector is included in this first version.

## Phase 3 · Cited answers

Cited answers for Chatbase: Retrieve matching passages with SQLite full-text search and supply only those passages to the local model.

### Steps

```sh
python -m unittest discover -s test
```

### Done when

- [ ] The return-policy answer cites an existing enabled chunk containing the policy.
- [ ] The unknown question produces an abstention, and hostile source instructions cannot trigger tools.

### Watch out

- A valid citation id alone does not prove entailment; keep an owner-reviewed answer evaluation set.

## Phase 4 · Widget and handoff

Widget and handoff for Chatbase: Build an iframe chat widget with bounded messages, signed conversation identity and explicit allowed embedding origins.

### Steps

```sh
python -m unittest discover -s test
```

### Done when

- [ ] One signed visitor cannot read another conversation by changing its id.
- [ ] Requesting a person creates one handoff and does not invent an order lookup result.

### Watch out

- Voice, write actions, customer account tools and omnichannel support are excluded; this FAQ agent cannot change orders.

## Phase 5 · Portable delivery and clean-start proof

Finish the Chatbase replacement with a documented first-run path and a recoverable export.

### Steps

```sh
python -m pip install -r requirements.txt
python -m unittest discover -s test
```

### Done when

- [ ] A new folder reaches the documented fixture outcome using only README instructions.
- [ ] A backup restored in isolation preserves the proof data and can be opened.

### Watch out

- Export without an import or restore rehearsal is not verified portability.

## Not in this build

## After v1, if you want it

===== .env.example =====
# Copy to .env and fill in. Never commit .env; this file documents it.

# Required. A new private folder this application creates on first run; never serve it as static files.
DATA_DIR=./data/chatbase

# Required · secret. Generate with python -c "import secrets; print(secrets.token_hex(32))" and save only in .env.
SECRET_KEY=replace-with-a-random-secret

# Required. Use indie for loopback-only personal work. Product mode must refuse to start until owner login and the deployment gates are configured.
APP_MODE=indie

# Required. The origin used for local forms and links; replace with the chosen HTTPS origin in product mode.
SITE_URL=http://127.0.0.1:5000

# Required. Keep the inference service bound to loopback; the backend calls it, never the public browser.
OLLAMA_URL=http://127.0.0.1:11434

# Required. Copy the exact non-cloud name from ollama list; preflight must reject this placeholder.
LOCAL_MODEL=replace-with-installed-local-model

You are building a production product version of Chatbase.

Create the following project files first, then implement the application by following them. Keep the files updated as decisions change. Do not collapse this into a single README or prompt.

===== PRODUCT.md =====
# Chatbase · product brief

## Problem

## Product outcome

An operated version of a bounded support FAQ assistant with citations and a human handoff queue, with authorized access, restore evidence and honest scope limits.

## Target user

A builder who needs a maintainable product foundation, not a one-off demo.

## Required capabilities

- Node.js 22, Docker, and PostgreSQL with pgvector
- an OpenAI API key for generation and embeddings
- object storage for uploaded source files
- a public HTTPS origin for the embeddable widget and webhooks

## Explicit non-goals for v1

## Success criteria

- Reimporting an unchanged source does not duplicate chunks.
- The return-policy answer cites an existing enabled chunk containing the policy.
- One signed visitor cannot read another conversation by changing its id.
- A private-data authorization check passes using a second identity.
- An off-host restore drill has a date, observed result and recovery time in OPERATIONS.md.
- README names the excluded vendor capabilities and actual provider/hosting costs.

===== BRIEF.md =====
# Build brief · Chatbase

The one-shot brief this plan expands. `BUILD_PLAN.md` (or `MILESTONES.md`) is the same sequence broken into steps and checks; where the two disagree, the plan wins.

Done when: Reimporting an unchanged source does not duplicate chunks. Disabling a source removes it from subsequent retrieval.
Do not build yet: Do not start cited answers until these checks pass.

### Phase 3 · Cited answers
Build: Cited answers for Chatbase: Retrieve matching passages with SQLite full-text search and supply only those passages to the local model.

### Phase 4 · Widget and handoff
Build: Widget and handoff for Chatbase: Build an iframe chat widget with bounded messages, signed conversation identity and explicit allowed embedding origins.

### Phase 5 · Portable delivery and clean-start proof
Build: Finish the Chatbase replacement with a documented first-run path and a recoverable export.

===== ARCHITECTURE.md =====
# Architecture · Chatbase

## Stack

## Modules

Each module has one owner concern and a documented way to replace it.

| Module | Owns | How to replace it |
| --- | --- | --- |
| Data and identity | sources(id,title,url,text,hash,enabled); chunks(id,source_id,text,position); conversations(id,visitor_hash); messages(id,conversation_id,role,text); citations(message_id,chunk_id); handoffs(id,conversation_id,email,summary,state) | Keep stable ids and a versioned repository interface; migrate storage behind it with a fixture round-trip. |
| Controlled knowledge intake | Accept owner-uploaded TXT/Markdown and pasted text with provenance; split into bounded paragraphs and index locally. | Keep the acceptance contract: Reimporting an unchanged source does not duplicate chunks. Replace only this module after replaying its fixture; preserve ids and existing records. |
| Cited answers | Retrieve matching passages with SQLite full-text search and supply only those passages to the local model. | Keep the acceptance contract: The return-policy answer cites an existing enabled chunk containing the policy. Replace only this module after replaying its fixture; preserve ids and existing records. |
| Widget and handoff | Build an iframe chat widget with bounded messages, signed conversation identity and explicit allowed embedding origins. | Keep the acceptance contract: One signed visitor cannot read another conversation by changing its id. Replace only this module after replaying its fixture; preserve ids and existing records. |
| Operations | Use sqlite3.Connection.backup for the database in DATA_DIR and pause writers while taking the matching attachment snapshot. Keep schema version, hashes and a manifest for chatbase. | Change host or backup destination only after an isolated restore and authorization check. |

## Configuration

Every runtime setting is an environment variable documented in `.env.example`, validated at startup, with a safe local default wherever one exists.

- `DATA_DIR` · required · A new private folder this application creates on first run; never serve it as static files.
- `SECRET_KEY` · required, secret · Generate with python -c "import secrets; print(secrets.token_hex(32))" and save only in .env.
- `APP_MODE` · required · Use indie for loopback-only personal work. Product mode must refuse to start until owner login and the deployment gates are configured.
- `SITE_URL` · required · The origin used for local forms and links; replace with the chosen HTTPS origin in product mode.
- `OLLAMA_URL` · required · Keep the inference service bound to loopback; the backend calls it, never the public browser.
- `LOCAL_MODEL` · required · Copy the exact non-cloud name from ollama list; preflight must reject this placeholder.

## Production baseline

- Security: least privilege, input validation at every boundary, secret redaction in logs, rate limits on abuse-prone paths, no invented security primitives.
- Data: explicit schema and migrations, transactional writes where integrity matters, backup and restore procedures that have been exercised.
- Integrations: adapters around third-party providers, idempotent webhook or job processing, bounded retries, timeouts.
- Observability: structured logs with request or operation ids, an error-tracking hook, and health and readiness checks where a server exists.
- Quality: unit tests for domain rules, integration tests at module boundaries, one end-to-end test of the critical path.

## Decision records

For each dependency in the stack table, keep a short note: why it was chosen, its failure mode, and how it is replaced. Do not add infrastructure until a requirement in `PRODUCT.md` justifies it.

===== AGENTS.md =====
# Agent instructions · Chatbase product build

- Read `PRODUCT.md` and `ARCHITECTURE.md` before changing code. The stack is fixed: Python 3.12+ and Flask, SQLite through Python sqlite3, Jinja HTML templates, CSS and vanilla JavaScript, Python unittest and browser acceptance fixtures, Ollama local HTTP API.
- Implement milestone by milestone from `MILESTONES.md`; keep each change reviewable and leave the application runnable at every commit.
- Treat authentication, payments, encryption, imports, webhooks and destructive actions as high-risk boundaries when present.
- Never invent cryptography or silently weaken a requirement to make a check pass.
- Put every external service behind an interface with a deterministic fake for tests.
- Add migrations and rollback or recovery notes for every persistent data change.
- Log useful operational context without credentials, tokens, passwords or personal data.
- Update documentation and run every check before completing a milestone.

## Known traps

===== MILESTONES.md =====
# Delivery milestones · Chatbase

Estimated effort: **multi-day** for the indie phases; the production-only milestones add the trust and operability layer.

## M1 · Local foundation and data contract

Create the runnable local foundation for a bounded support FAQ assistant with citations and a human handoff queue, including the real schema and acceptance fixture.

### Steps

```sh
python -m venv .venv
python -m pip install Flask Flask-WTF python-dotenv waitress tzdata
python -m pip freeze > requirements.txt
python -m flask --app app run --host 127.0.0.1
```

```sh
python manage.py init
python manage.py seed
```

```sh
python scripts/model_probe.py
```

### Done when

### Watch out

- The Flask development server is for local development only. Loopback binding alone does not replace CSRF and origin validation.

## M2 · Controlled knowledge intake

Controlled knowledge intake for Chatbase: Accept owner-uploaded TXT/Markdown and pasted text with provenance; split into bounded paragraphs and index locally.

### Steps

```sh
python -m unittest discover -s test
```

### Done when

- [ ] Reimporting an unchanged source does not duplicate chunks.
- [ ] Disabling a source removes it from subsequent retrieval.

### Watch out

- No open-web crawler, document execution or automatic connector is included in this first version.

## M3 · Cited answers

Cited answers for Chatbase: Retrieve matching passages with SQLite full-text search and supply only those passages to the local model.

### Steps

```sh
python -m unittest discover -s test
```

### Done when

- [ ] The return-policy answer cites an existing enabled chunk containing the policy.
- [ ] The unknown question produces an abstention, and hostile source instructions cannot trigger tools.

### Watch out

- A valid citation id alone does not prove entailment; keep an owner-reviewed answer evaluation set.

## M4 · Widget and handoff

Widget and handoff for Chatbase: Build an iframe chat widget with bounded messages, signed conversation identity and explicit allowed embedding origins.

### Steps

```sh
python -m unittest discover -s test
```

### Done when

- [ ] One signed visitor cannot read another conversation by changing its id.
- [ ] Requesting a person creates one handoff and does not invent an order lookup result.

### Watch out

- Voice, write actions, customer account tools and omnichannel support are excluded; this FAQ agent cannot change orders.

## M5 · Portable delivery and clean-start proof

Finish the Chatbase replacement with a documented first-run path and a recoverable export.

### Steps

```sh
python -m pip install -r requirements.txt
python -m unittest discover -s test
```

### Done when

- [ ] A new folder reaches the documented fixture outcome using only README instructions.
- [ ] A backup restored in isolation preserves the proof data and can be opened.

### Watch out

- Export without an import or restore rehearsal is not verified portability.

## M6 · Operate and share with access controls (production only)

Product-builder only: make a bounded support FAQ assistant with citations and a human handoff queue operable for its stated audience before allowing remote access.

### Steps

1. Establish owner access before exposure
Add owner/member/viewer authorization as appropriate, password hashes through Werkzeug, expiring server-side sessions, CSRF and login rate limits. Add users, sessions and audit schema migrations plus an owner-init command prompting securely for its password. Enforce ownership on every private read/write/download and fail closed when product setup is incomplete. Existing learner or admin identities must use this same boundary.
Files: `ARCHITECTURE.md`, `OPERATIONS.md`
2. Deploy a monitored service
Serve app:app with Waitress on loopback behind Caddy HTTPS. Configure trusted proxy headers narrowly, Secure/HttpOnly/SameSite cookies, upload limits, request ids and structured logs without private content. Run background jobs as a separate supervised process when the domain workflow requires them.
Files: `Caddyfile`, `OPERATIONS.md`

```sh
waitress-serve --listen=127.0.0.1:5000 app:app
```

3. Rehearse loss and recovery
Add a scheduled encrypted off-host backup using this procedure: Use sqlite3.Connection.backup for the database in DATA_DIR and pause writers while taking the matching attachment snapshot. Keep schema version, hashes and a manifest for chatbase. Monitor health, disk space, failed jobs and backup age. Stop the service to test an alert; rehearse recovery with outbound effects disabled and record recovery time. Rotate any exposed credentials and require reconciliation before resuming queues.
Files: `OPERATIONS.md`, `release-checklist.md`

### Done when

- [ ] An unauthenticated visitor cannot read private records, download private files or mutate the admin surface.
- [ ] A viewer cannot mutate data by making a direct request, and guessed ids cannot cross ownership boundaries.
- [ ] Stopping the service triggers the configured alert and the latest off-host backup restores the fixture in isolation.

### Watch out

- Product mode is additional engineering, not a switch that makes a local prototype production-ready.

===== OPERATIONS.md =====
# Operations · Chatbase

## Backup

Use sqlite3.Connection.backup for the database in DATA_DIR and pause writers while taking the matching attachment snapshot. Keep schema version, hashes and a manifest for chatbase.

## Restore

Restore into a separate empty directory or isolated deployment, with all outbound actions and schedules disabled. Recreate this proof: Five supplied FAQ documents, an answerable return-policy question, an unknown question and a source containing hostile instructions. Compare ids/counts and open the result before recording the restore date. Never test a restore over the only live copy.

Do a restore drill before the first real user, and write the date here when it passes.

## Monitoring

Check HTTP availability, free disk, backup age and domain failures for Controlled knowledge intake, Cited answers, Widget and handoff. Alert without logging private payloads.

## Incident checklist

Pause outbound jobs and public intake as appropriate, preserve redacted diagnostics, restore an isolated verified backup, rotate affected credentials and reconcile uncertain effects before reopening.

1. Contain the issue without destroying evidence or user data.
2. Record the timeline and affected scope.
3. Rotate exposed secrets and revoke compromised sessions or credentials.
4. Restore from a verified backup when needed.
5. Document the root cause, the remediation and the regression test.

## Release gate

- [ ] Reimporting an unchanged source does not duplicate chunks.
- [ ] The return-policy answer cites an existing enabled chunk containing the policy.
- [ ] One signed visitor cannot read another conversation by changing its id.
- [ ] A private-data authorization check passes using a second identity.
- [ ] An off-host restore drill has a date, observed result and recovery time in OPERATIONS.md.
- [ ] README names the excluded vendor capabilities and actual provider/hosting costs.

## Launch constraint

Do not market omitted Chatbase capabilities as implemented. The non-goals in `PRODUCT.md` remain user-visible limitations until they are deliberately delivered.

===== .env.example =====
# Copy to .env and fill in. Never commit .env; this file documents it.

# Required. A new private folder this application creates on first run; never serve it as static files.
DATA_DIR=./data/chatbase

# Required · secret. Generate with python -c "import secrets; print(secrets.token_hex(32))" and save only in .env.
SECRET_KEY=replace-with-a-random-secret

# Required. Use indie for loopback-only personal work. Product mode must refuse to start until owner login and the deployment gates are configured.
APP_MODE=indie

# Required. The origin used for local forms and links; replace with the chosen HTTPS origin in product mode.
SITE_URL=http://127.0.0.1:5000

# Required. Keep the inference service bound to loopback; the backend calls it, never the public browser.
OLLAMA_URL=http://127.0.0.1:11434

# Required. Copy the exact non-cloud name from ollama list; preflight must reject this placeholder.
LOCAL_MODEL=replace-with-installed-local-model

README.md

# Chatbase · indie build

Build a bounded support FAQ assistant with citations and a human handoff queue.

Estimated effort: **multi-day**. Work `BUILD_PLAN.md` top to bottom · every phase ends in a check that has to pass before the next one starts.

## Stack

## Before you start

Have every one of these ready. The plan assumes them from step one.

## Quick start

```sh
python -m venv .venv
python -m pip install Flask Flask-WTF python-dotenv waitress tzdata
python -m pip freeze > requirements.txt
python -m flask --app app run --host 127.0.0.1
```

Then copy `.env.example` to `.env` and fill in the values it documents.

## Honest limits

This build deliberately does not replace:

If one of those is essential to you, that is the reason to keep paying for Chatbase, and the README should say so rather than pretend.

BRIEF.md

# Build brief · Chatbase

The one-shot brief this plan expands. `BUILD_PLAN.md` (or `MILESTONES.md`) is the same sequence broken into steps and checks; where the two disagree, the plan wins.

Done when: Reimporting an unchanged source does not duplicate chunks. Disabling a source removes it from subsequent retrieval.
Do not build yet: Do not start cited answers until these checks pass.

### Phase 3 · Cited answers
Build: Cited answers for Chatbase: Retrieve matching passages with SQLite full-text search and supply only those passages to the local model.

### Phase 4 · Widget and handoff
Build: Widget and handoff for Chatbase: Build an iframe chat widget with bounded messages, signed conversation identity and explicit allowed embedding origins.

### Phase 5 · Portable delivery and clean-start proof
Build: Finish the Chatbase replacement with a documented first-run path and a recoverable export.

AGENTS.md

# Agent instructions · Chatbase indie build

## Known traps

BUILD_PLAN.md

# Build plan · Chatbase

Build a bounded support FAQ assistant with citations and a human handoff queue.

Phases are in dependency order. Each ends in a "Done when" list; treat an unticked item as a blocker, not a note.

## Phase 1 · Local foundation and data contract

Create the runnable local foundation for a bounded support FAQ assistant with citations and a human handoff queue, including the real schema and acceptance fixture.

### Steps

```sh
python -m venv .venv
python -m pip install Flask Flask-WTF python-dotenv waitress tzdata
python -m pip freeze > requirements.txt
python -m flask --app app run --host 127.0.0.1
```

```sh
python manage.py init
python manage.py seed
```

```sh
python scripts/model_probe.py
```

### Done when

### Watch out

- The Flask development server is for local development only. Loopback binding alone does not replace CSRF and origin validation.

## Phase 2 · Controlled knowledge intake

Controlled knowledge intake for Chatbase: Accept owner-uploaded TXT/Markdown and pasted text with provenance; split into bounded paragraphs and index locally.

### Steps

```sh
python -m unittest discover -s test
```

### Done when

- [ ] Reimporting an unchanged source does not duplicate chunks.
- [ ] Disabling a source removes it from subsequent retrieval.

### Watch out

- No open-web crawler, document execution or automatic connector is included in this first version.

## Phase 3 · Cited answers

Cited answers for Chatbase: Retrieve matching passages with SQLite full-text search and supply only those passages to the local model.

### Steps

```sh
python -m unittest discover -s test
```

### Done when

- [ ] The return-policy answer cites an existing enabled chunk containing the policy.
- [ ] The unknown question produces an abstention, and hostile source instructions cannot trigger tools.

### Watch out

- A valid citation id alone does not prove entailment; keep an owner-reviewed answer evaluation set.

## Phase 4 · Widget and handoff

Widget and handoff for Chatbase: Build an iframe chat widget with bounded messages, signed conversation identity and explicit allowed embedding origins.

### Steps

```sh
python -m unittest discover -s test
```

### Done when

- [ ] One signed visitor cannot read another conversation by changing its id.
- [ ] Requesting a person creates one handoff and does not invent an order lookup result.

### Watch out

- Voice, write actions, customer account tools and omnichannel support are excluded; this FAQ agent cannot change orders.

## Phase 5 · Portable delivery and clean-start proof

Finish the Chatbase replacement with a documented first-run path and a recoverable export.

### Steps

```sh
python -m pip install -r requirements.txt
python -m unittest discover -s test
```

### Done when

- [ ] A new folder reaches the documented fixture outcome using only README instructions.
- [ ] A backup restored in isolation preserves the proof data and can be opened.

### Watch out

- Export without an import or restore rehearsal is not verified portability.

## Not in this build

## After v1, if you want it

.env.example

# Copy to .env and fill in. Never commit .env; this file documents it.

# Required. A new private folder this application creates on first run; never serve it as static files.
DATA_DIR=./data/chatbase

# Required · secret. Generate with python -c "import secrets; print(secrets.token_hex(32))" and save only in .env.
SECRET_KEY=replace-with-a-random-secret

# Required. Use indie for loopback-only personal work. Product mode must refuse to start until owner login and the deployment gates are configured.
APP_MODE=indie

# Required. The origin used for local forms and links; replace with the chosen HTTPS origin in product mode.
SITE_URL=http://127.0.0.1:5000

# Required. Keep the inference service bound to loopback; the backend calls it, never the public browser.
OLLAMA_URL=http://127.0.0.1:11434

# Required. Copy the exact non-cloud name from ollama list; preflight must reject this placeholder.
LOCAL_MODEL=replace-with-installed-local-model

PRODUCT.md

# Chatbase · product brief

## Problem

A useful web version is a serious multi-day build, not magic: crawl or upload sources, retrieve tenant-scoped passages, stream cited answers, run one guarded business action, and hand uncertain requests to a person. Rebuilding Chatbase itself is a different project. Its paid surface combines source management and retraining, procedures and custom UI actions, a polished widget, helpdesk, identity, analytics, voice, email, social and CRM channels, plus the reliability and security work that keeps an agent safe in front of customers.

## Product outcome

An operated version of a bounded support FAQ assistant with citations and a human handoff queue, with authorized access, restore evidence and honest scope limits.

## Target user

A builder who needs a maintainable product foundation, not a one-off demo.

## Required capabilities

- Node.js 22, Docker, and PostgreSQL with pgvector
- an OpenAI API key for generation and embeddings
- object storage for uploaded source files
- a public HTTPS origin for the embeddable widget and webhooks

## Explicit non-goals for v1

- Voice, write actions, customer account tools and omnichannel support are excluded; this FAQ agent cannot change orders.
- This is a bounded support FAQ assistant with citations and a human handoff queue; full Chatbase feature parity, vendor hosting and support are not included.
- the no-code agent lifecycle: many managed source connectors, background retraining, source suggestions, and production ingestion diagnostics
- the channel network: website chat, email, voice and telephony, Slack, WhatsApp, Messenger, Instagram, Shopify, WordPress, Zapier, Zendesk, Salesforce, and other integrations
- the support operation around the agent: a team helpdesk, tickets, assignment, contact identity, authenticated personalization, escalation, and access controls
- the action and procedure builder: server, client, button, and custom widget actions with testing, permissions, and integration-specific behavior
- the production layer: topic and sentiment analytics, review workflows, abuse controls, delivery retries, observability, data governance, compliance, and support

## Success criteria

- Reimporting an unchanged source does not duplicate chunks.
- The return-policy answer cites an existing enabled chunk containing the policy.
- One signed visitor cannot read another conversation by changing its id.
- A private-data authorization check passes using a second identity.
- An off-host restore drill has a date, observed result and recovery time in OPERATIONS.md.
- README names the excluded vendor capabilities and actual provider/hosting costs.

BRIEF.md

# Build brief · Chatbase

The one-shot brief this plan expands. `BUILD_PLAN.md` (or `MILESTONES.md`) is the same sequence broken into steps and checks; where the two disagree, the plan wins.

Done when: Reimporting an unchanged source does not duplicate chunks. Disabling a source removes it from subsequent retrieval.
Do not build yet: Do not start cited answers until these checks pass.

### Phase 3 · Cited answers
Build: Cited answers for Chatbase: Retrieve matching passages with SQLite full-text search and supply only those passages to the local model.

### Phase 4 · Widget and handoff
Build: Widget and handoff for Chatbase: Build an iframe chat widget with bounded messages, signed conversation identity and explicit allowed embedding origins.

### Phase 5 · Portable delivery and clean-start proof
Build: Finish the Chatbase replacement with a documented first-run path and a recoverable export.

ARCHITECTURE.md

# Architecture · Chatbase

## Stack

| Part | Choice | Why |
| --- | --- | --- |
| Runtime | Python 3.12+ and Flask | A small local HTTP application with readable server-side handlers and no JavaScript build pipeline. |
| Storage | SQLite through Python sqlite3 | Local ownership with explicit transactions and a portable backup path. |
| Interface | Jinja HTML templates, CSS and vanilla JavaScript | Implement the specified screens directly with labelled controls and progressive enhancement. |
| Verification | Python unittest and browser acceptance fixtures | Check invariants and the actual user workflow before the next phase. |
| Generation | Ollama local HTTP API | A backend-only adapter with a configurable installed local model and visible failure states. |

## Modules

Each module has one owner concern and a documented way to replace it.

| Module | Owns | How to replace it |
| --- | --- | --- |
| Data and identity | sources(id,title,url,text,hash,enabled); chunks(id,source_id,text,position); conversations(id,visitor_hash); messages(id,conversation_id,role,text); citations(message_id,chunk_id); handoffs(id,conversation_id,email,summary,state) | Keep stable ids and a versioned repository interface; migrate storage behind it with a fixture round-trip. |
| Controlled knowledge intake | Accept owner-uploaded TXT/Markdown and pasted text with provenance; split into bounded paragraphs and index locally. | Keep the acceptance contract: Reimporting an unchanged source does not duplicate chunks. Replace only this module after replaying its fixture; preserve ids and existing records. |
| Cited answers | Retrieve matching passages with SQLite full-text search and supply only those passages to the local model. | Keep the acceptance contract: The return-policy answer cites an existing enabled chunk containing the policy. Replace only this module after replaying its fixture; preserve ids and existing records. |
| Widget and handoff | Build an iframe chat widget with bounded messages, signed conversation identity and explicit allowed embedding origins. | Keep the acceptance contract: One signed visitor cannot read another conversation by changing its id. Replace only this module after replaying its fixture; preserve ids and existing records. |
| Operations | Use sqlite3.Connection.backup for the database in DATA_DIR and pause writers while taking the matching attachment snapshot. Keep schema version, hashes and a manifest for chatbase. | Change host or backup destination only after an isolated restore and authorization check. |

## Configuration

Every runtime setting is an environment variable documented in `.env.example`, validated at startup, with a safe local default wherever one exists.

- `DATA_DIR` · required · A new private folder this application creates on first run; never serve it as static files.
- `SECRET_KEY` · required, secret · Generate with python -c "import secrets; print(secrets.token_hex(32))" and save only in .env.
- `APP_MODE` · required · Use indie for loopback-only personal work. Product mode must refuse to start until owner login and the deployment gates are configured.
- `SITE_URL` · required · The origin used for local forms and links; replace with the chosen HTTPS origin in product mode.
- `OLLAMA_URL` · required · Keep the inference service bound to loopback; the backend calls it, never the public browser.
- `LOCAL_MODEL` · required · Copy the exact non-cloud name from ollama list; preflight must reject this placeholder.

## Production baseline

- Security: least privilege, input validation at every boundary, secret redaction in logs, rate limits on abuse-prone paths, no invented security primitives.
- Data: explicit schema and migrations, transactional writes where integrity matters, backup and restore procedures that have been exercised.
- Integrations: adapters around third-party providers, idempotent webhook or job processing, bounded retries, timeouts.
- Observability: structured logs with request or operation ids, an error-tracking hook, and health and readiness checks where a server exists.
- Quality: unit tests for domain rules, integration tests at module boundaries, one end-to-end test of the critical path.

## Decision records

For each dependency in the stack table, keep a short note: why it was chosen, its failure mode, and how it is replaced. Do not add infrastructure until a requirement in `PRODUCT.md` justifies it.

AGENTS.md

# Agent instructions · Chatbase product build

## Known traps

MILESTONES.md

# Delivery milestones · Chatbase

Estimated effort: **multi-day** for the indie phases; the production-only milestones add the trust and operability layer.

## M1 · Local foundation and data contract

Create the runnable local foundation for a bounded support FAQ assistant with citations and a human handoff queue, including the real schema and acceptance fixture.

### Steps

```sh
python -m venv .venv
python -m pip install Flask Flask-WTF python-dotenv waitress tzdata
python -m pip freeze > requirements.txt
python -m flask --app app run --host 127.0.0.1
```

```sh
python manage.py init
python manage.py seed
```

```sh
python scripts/model_probe.py
```

### Done when

### Watch out

- The Flask development server is for local development only. Loopback binding alone does not replace CSRF and origin validation.

## M2 · Controlled knowledge intake

Controlled knowledge intake for Chatbase: Accept owner-uploaded TXT/Markdown and pasted text with provenance; split into bounded paragraphs and index locally.

### Steps

```sh
python -m unittest discover -s test
```

### Done when

- [ ] Reimporting an unchanged source does not duplicate chunks.
- [ ] Disabling a source removes it from subsequent retrieval.

### Watch out

- No open-web crawler, document execution or automatic connector is included in this first version.

## M3 · Cited answers

Cited answers for Chatbase: Retrieve matching passages with SQLite full-text search and supply only those passages to the local model.

### Steps

```sh
python -m unittest discover -s test
```

### Done when

- [ ] The return-policy answer cites an existing enabled chunk containing the policy.
- [ ] The unknown question produces an abstention, and hostile source instructions cannot trigger tools.

### Watch out

- A valid citation id alone does not prove entailment; keep an owner-reviewed answer evaluation set.

## M4 · Widget and handoff

Widget and handoff for Chatbase: Build an iframe chat widget with bounded messages, signed conversation identity and explicit allowed embedding origins.

### Steps

```sh
python -m unittest discover -s test
```

### Done when

- [ ] One signed visitor cannot read another conversation by changing its id.
- [ ] Requesting a person creates one handoff and does not invent an order lookup result.

### Watch out

- Voice, write actions, customer account tools and omnichannel support are excluded; this FAQ agent cannot change orders.

## M5 · Portable delivery and clean-start proof

Finish the Chatbase replacement with a documented first-run path and a recoverable export.

### Steps

```sh
python -m pip install -r requirements.txt
python -m unittest discover -s test
```

### Done when

- [ ] A new folder reaches the documented fixture outcome using only README instructions.
- [ ] A backup restored in isolation preserves the proof data and can be opened.

### Watch out

- Export without an import or restore rehearsal is not verified portability.

## M6 · Operate and share with access controls (production only)

Product-builder only: make a bounded support FAQ assistant with citations and a human handoff queue operable for its stated audience before allowing remote access.

### Steps

```sh
waitress-serve --listen=127.0.0.1:5000 app:app
```

### Done when

### Watch out

- Product mode is additional engineering, not a switch that makes a local prototype production-ready.

OPERATIONS.md

# Operations · Chatbase

## Backup

Use sqlite3.Connection.backup for the database in DATA_DIR and pause writers while taking the matching attachment snapshot. Keep schema version, hashes and a manifest for chatbase.

## Restore

Do a restore drill before the first real user, and write the date here when it passes.

## Monitoring

Check HTTP availability, free disk, backup age and domain failures for Controlled knowledge intake, Cited answers, Widget and handoff. Alert without logging private payloads.

## Incident checklist

Pause outbound jobs and public intake as appropriate, preserve redacted diagnostics, restore an isolated verified backup, rotate affected credentials and reconcile uncertain effects before reopening.

## Release gate

## Launch constraint

Do not market omitted Chatbase capabilities as implemented. The non-goals in `PRODUCT.md` remain user-visible limitations until they are deliberately delivered.

.env.example

# Copy to .env and fill in. Never commit .env; this file documents it.

# Required. A new private folder this application creates on first run; never serve it as static files.
DATA_DIR=./data/chatbase

# Required · secret. Generate with python -c "import secrets; print(secrets.token_hex(32))" and save only in .env.
SECRET_KEY=replace-with-a-random-secret

# Required. Use indie for loopback-only personal work. Product mode must refuse to start until owner login and the deployment gates are configured.
APP_MODE=indie

# Required. The origin used for local forms and links; replace with the chosen HTTPS origin in product mode.
SITE_URL=http://127.0.0.1:5000

# Required. Keep the inference service bound to loopback; the backend calls it, never the public browser.
OLLAMA_URL=http://127.0.0.1:11434

# Required. Copy the exact non-cloud name from ollama list; preflight must reject this placeholder.
LOCAL_MODEL=replace-with-installed-local-model

$ choose a build depth, inspect the files, then open the complete pack in your agent

why people still pay

The chat bubble and one RAG route are only the visible edge. The subscription buys the control plane around them: continuously managed sources, safe actions, customer identity, channel adapters, a human inbox, analytics, and someone else owning delivery and model regressions. Building is reasonable when one web agent and full data control are enough; paying is reasonable when the agent is part of a real support operation.

what you lose

xthe no-code agent lifecycle: many managed source connectors, background retraining, source suggestions, and production ingestion diagnostics

xthe channel network: website chat, email, voice and telephony, Slack, WhatsApp, Messenger, Instagram, Shopify, WordPress, Zapier, Zendesk, Salesforce, and other integrations

xthe support operation around the agent: a team helpdesk, tickets, assignment, contact identity, authenticated personalization, escalation, and access controls

xthe action and procedure builder: server, client, button, and custom widget actions with testing, permissions, and integration-specific behavior

xthe production layer: topic and sentiment analytics, review workflows, abuse controls, delivery retries, observability, data governance, compliance, and support

share on X ↗"I just replaced Chatbase ($150/mo) with one prompt"

the escape hatch

Don't feel like building it? These folks already made it free.

TiledeskA self-hosted RAG support agent across chat, email and voice; the free software arrives as an infrastructure diagram.316jul 2026open source↗

no votes, no pay-to-list · just what's real

the numbers

Chatbase pricing

plan	monthly	annual (per mo)	what you get
free	$0/workspace	$0/workspace	1 agent, 50 message credits/month, 1 MB training content, 1 member and 0 AI actions/agent
hobby	$40/workspace	$32/workspace	1 agent, 700 message credits/month, 5 AI actions/agent, 10 MB training content and 2 members
standard	$150/workspace	$120/workspace	1 agent, 4,000 message credits/month, 8 AI actions/agent, 20 MB training content, 3 members and 10 concurrent voice calls
pro	$500/workspace	$400/workspace	1 agent, 15,000 message credits/month, 12 AI actions/agent, 40 MB training content, 5 members and 20 concurrent voice calls
enterprise	custom	—	Custom higher limits, roles, SSO, white-labeling, audit logs, SLA and zero-data-retention options

free tier1 agent; 50 message credits/month; 1 MB training data; 1 member; agents deleted after 14 inactive days

billingmonthly + annual (-20%); paid plans have a 7-day trial

hidden costsAuto-recharge message credits cost $40 per 1,000 and do not expire; each extra agent costs $300/year; removing Chatbase branding costs $1,188/year.

verified 2026-08-12 · source ↗

questions

Vibecode Chatbase

Kinda. The core of Chatbase is buildable in a weekend with the prompt on this page, but there are real gaps: the no-code agent lifecycle: many managed source connectors, background retraining, source suggestions, and production ingestion diagnostics, the channel network: website chat, email, voice and telephony, Slack, WhatsApp, Messenger, Instagram, Shopify, WordPress, Zapier, Zendesk, Salesforce, and other integrations. Read the honest list above before committing.

How much does Chatbase cost?

Chatbase costs about $150/month (Standard, checked 2026-08-01), which is $1800 per year.

What do I lose by replacing Chatbase?

Honestly: the no-code agent lifecycle: many managed source connectors, background retraining, source suggestions, and production ingestion diagnostics; the channel network: website chat, email, voice and telephony, Slack, WhatsApp, Messenger, Instagram, Shopify, WordPress, Zapier, Zendesk, Salesforce, and other integrations; the support operation around the agent: a team helpdesk, tickets, assignment, contact identity, authenticated personalization, escalation, and access controls; the action and procedure builder: server, client, button, and custom widget actions with testing, permissions, and integration-specific behavior; the production layer: topic and sentiment analytics, review workflows, abuse controls, delivery retries, observability, data governance, compliance, and support. If any of those are load-bearing for you, keep paying.

Is there an open-source alternative to Chatbase?

Yes: Tiledesk (A self-hosted RAG support agent across chat, email and voice; the free software arrives as an infrastructure diagram.) The prompt is for when you want it exactly your way.