Angler

Semantic search for your company's content. Connect your markdown repos, Notion workspace, Google Docs, and Dropbox — then search everything with a single API.

What it does

Angler pulls content from multiple sources, chunks it, generates embeddings, and serves a search API powered by ChromaDB. It finds content by meaning, not just keywords — so searching "what did we discuss about pricing" will find relevant passages even if they never use the word "pricing."

Sources

Source	What it indexes	Auth
Git	Markdown files from a git repo	GitHub token (for private repos)
Notion	Pages and database entries	Internal integration token
Google Docs	Documents in a Drive folder	Service account credentials
Dropbox	Text files in a Dropbox folder	Access token
SQLite	Rows from SQLite tables	File path to .db

Quick start

Option 1: Config file

Copy angler.yaml.example to angler.yaml and enable the sources you need:

sources:
  - type: git
    repo_url: https://github.com/yourorg/docs.git

  - type: notion
    # token set via NOTION_TOKEN env var

  - type: google_docs
    folder_id: "1a2b3c..."

  - type: dropbox
    # token set via DROPBOX_TOKEN env var
    folder_path: "/Documents"  # optional, "" = root

  - type: sqlite
    db_path: "~/agency/agency.db"
    tables:
      - table: employees
        text_columns: [name, role, notes]
        id_column: id
        updated_column: updated_at

pip install -r requirements.txt
uvicorn app:app --reload

Option 2: Environment variables

For simple deployments, skip the config file and use env vars:

Variable	Source	Description
REPO_URL	Git	URL of your markdown repo
GITHUB_TOKEN	Git	For private GitHub repos
NOTION_TOKEN	Notion	Internal integration token
NOTION_DATABASES	Notion	Comma-separated database IDs (optional)
GOOGLE_DOCS_FOLDER_ID	Google Docs	Root folder to index
GOOGLE_CREDENTIALS_JSON	Google Docs	Service account JSON
DROPBOX_TOKEN	Dropbox	App access token
DROPBOX_FOLDER_PATH	Dropbox	Folder to index (optional, default: root)
SQLITE_DB_PATH	SQLite	Path to .db file
SQLITE_TABLES	SQLite	JSON array of table configs (see below)
CHROMA_DIR	—	Where to persist the index (default: /data/chroma)
PORT	—	Server port (default: 8000)

Deploy to Railway

1. Fork this repo
2. Connect it to Railway
3. Set your source env vars
4. Railway will use the included Dockerfile and railway.toml

API

GET /search?q=your+query&n=5

Search your content. Returns ranked results with relevance scores.

{
  "query": "your query",
  "results": [
    {
      "id": "git://docs/meeting-notes.md::chunk_3",
      "score": 0.82,
      "content": "...matching text...",
      "metadata": {
        "source": "git",
        "path": "docs/meeting-notes.md",
        "title": "Meeting Notes"
      }
    }
  ]
}

Filter by source type:

GET /search?q=pricing+discussion&source=notion

POST /reindex

Re-fetch all sources and rebuild the index.

GET /health

Check service status, configured sources, and index size.

Source setup

Git

Just set REPO_URL. For private repos, also set GITHUB_TOKEN.

Notion

1. Create an internal integration at notion.so/my-integrations
2. Copy the integration token
3. Share your top-level workspace pages with the integration (this grants access to all child pages)
4. Set NOTION_TOKEN

Google Docs

1. Create a service account in Google Cloud Console
2. Enable the Google Drive and Google Docs APIs
3. Share your target folder with the service account's email address
4. Set GOOGLE_DOCS_FOLDER_ID and GOOGLE_CREDENTIALS_JSON

Dropbox

1. Create an app at dropbox.com/developers/apps
2. Generate an access token
3. Set DROPBOX_TOKEN and optionally DROPBOX_FOLDER_PATH

By default it indexes .md, .txt, .csv, .json, .yaml, .xml, .html, and .rtf files. You can customize this with the extensions config option.

SQLite

Point Angler at any SQLite database and tell it which tables and columns to index:

- type: sqlite
  db_path: "~/agency/agency.db"
  tables:
    - table: employees
      text_columns: [name, role, notes]
      id_column: id            # default: rowid
      updated_column: updated_at  # optional, enables incremental indexing
    - table: projects
      text_columns: [name, notes]
      id_column: id

Each row becomes a document with the text columns concatenated. The updated_column is optional — if provided, incremental reindexing will only fetch rows modified since the last run.

Using with Claude Code

Create a skill at ~/.claude/skills/search-docs/SKILL.md:

---
name: search-docs
description: Search our team's docs, notes, and knowledge base.
user_invocable: true
---

Use the Bash tool to query the search API:

\```bash
curl -s "https://your-deployment.up.railway.app/search?q=QUERY&n=5" | python3 -m json.tool
\```

Then use /search-docs in Claude Code to search your content naturally.

License

MIT