Pre-meeting briefs powered by Airweave. Pulls context from your CRM, docs, and codebase so you walk into every call prepared.

Sales reps waste time scrambling for context before calls. The information exists, scattered across Notion pages, Google Drive docs, and GitHub repos, but finding it takes too long.

This agent uses Airweave to search across all your connected sources at once and synthesize the results into a structured pre-meeting brief.

Briefing Request --> [Decomposition] --> [Source Search] --> [Synthesis] --> Account Brief
                          |                   |                  |
                          +-- Breaks into     +-- Airweave       +-- LLM synthesizes
                              targeted            searches           into structured
                              queries             Notion, Drive,     brief with
                              per source          GitHub             talking points

Example: "Prep me for the Acme Analytics call" --> 6 targeted queries --> 8 relevant docs --> structured brief with 5 sections

The frontend provides a visual walkthrough of the pipeline. It works out of the box with sample data -- no API keys needed.

# Clone and setup
git clone https://github.com/airweave-ai/sales-context-agent.git
cd sales-context-agent
cp .env.example .env

# Optional: Add LLM key for smarter query decomposition (works without it too)
# ANTHROPIC_API_KEY=your_key  (or OPENAI_API_KEY)

# Start backend
cd backend
python -m venv venv && source venv/bin/activate
pip install -r requirements.txt
uvicorn main:app --reload --port 8000

# Start frontend (new terminal)
cd frontend
npm install && npm run dev

Open http://localhost:3000 and click Generate Brief.

The interactive visualization walks you through:

1. Briefing Request - Account name, meeting topic, and attendees
2. Query Decomposition - Request broken into targeted queries tagged by source (Notion, Google Drive, GitHub)
3. Source Search - Results from each source with relevance scores, found via Airweave
4. Context Synthesis - Search results combined into a structured brief with 5 sections
5. Brief Output - Full rendered brief with copy-to-clipboard and optional Slack posting

The demo uses:

• Sample data by default (realistic brief for "Acme Analytics" at PostHog)
• Mock search results showing what Airweave would return
• Preview mode for Slack (no actual messages sent)

To use this for real pre-meeting prep:

This is the core of the agent -- Airweave searches your Notion, Google Drive, GitHub, and 50+ other sources.

AIRWEAVE_API_KEY=your_key
AIRWEAVE_COLLECTION_ID=your_collection
AIRWEAVE_API_URL=https://api.airweave.ai  # or self-hosted

Set up your Airweave collection:

1. Create a collection at airweave.ai
2. Connect your sources (Notion, Google Drive, GitHub, Slack, etc.)
3. Wait for initial sync to complete

Now the agent will search your actual docs and repos when generating briefs.

Used for query decomposition and brief synthesis.

# Option 1: Anthropic (recommended)
ANTHROPIC_API_KEY=your_key
# ANTHROPIC_MODEL=claude-sonnet-4-20250514

# Option 2: OpenAI
OPENAI_API_KEY=your_key
# OPENAI_MODEL=gpt-4o

Post generated briefs directly to a Slack channel.

SLACK_ENABLED=true
SLACK_BOT_TOKEN=xoxb-your-token
SLACK_CHANNEL_ID=C0123456789

uvicorn main:app --host 0.0.0.0 --port 8000

# Trigger a brief via API
curl -X POST http://localhost:8000/api/run \
  -H "Content-Type: application/json" \
  -d '{"use_sample_data": false, "briefing_request": {"account": "Acme Corp", "topic": "Q1 renewal", "attendees": ["Jane Smith"]}}'

import asyncio
from main import run_pipeline, PipelineConfig

async def main():
    config = PipelineConfig(
        use_sample_data=False,
        briefing_request={
            "account": "Acme Corp",
            "topic": "Q1 renewal discussion",
            "attendees": ["Jane Smith", "Bob Johnson"],
        },
        post_to_slack=True,
    )
    result = await run_pipeline(config)
    print(f"Pipeline completed: {result.status}")

if __name__ == "__main__":
    asyncio.run(main())

Every generated brief includes:

• Account Snapshot - Company overview, tier, engagement history
• Relationship History - Past interactions, key contacts, deal context
• Open Concerns - Active issues, support tickets, blockers
• Product Updates - Recent changes relevant to this account
• Talking Points - Specific items to raise in the meeting

The agent uses Airweave's source filtering to search the right tool for each query:

from airweave import AirweaveSDK

client = AirweaveSDK(api_key=AIRWEAVE_API_KEY)

# Search Notion for account context
results = client.search.search(
    collection_id=COLLECTION_ID,
    query="Acme Analytics account overview",
    source_name="notion",
    limit=5
)

# Search GitHub for technical details
results = client.search.search(
    collection_id=COLLECTION_ID,
    query="Acme custom integration",
    source_name="github",
    limit=5
)

├── backend/
│   ├── main.py              # FastAPI + WebSocket + Pipeline orchestration
│   ├── config.py            # Configuration management
│   ├── schemas.py           # Pydantic data models
│   ├── samples/             # Sample briefing data for demo
│   ├── clients/             # External APIs (Airweave, Slack)
│   └── pipeline/
│       ├── decomposition.py # Query decomposition with source tagging
│       ├── search.py        # Airweave cross-source search
│       ├── synthesis.py     # LLM-powered brief generation
│       └── output.py        # Brief output + Slack posting
├── frontend/                # React demo visualization
│   ├── src/
│   │   ├── App.tsx          # Main app with WebSocket connection
│   │   └── components/      # Pipeline visualization components
└── .env.example             # All available environment variables

Connect to /ws/pipeline for real-time pipeline updates (used by the frontend):

const ws = new WebSocket('ws://localhost:8000/ws/pipeline');
ws.send(JSON.stringify({ type: 'run', config: { use_sample_data: true } }));
ws.onmessage = (e) => {
  const event = JSON.parse(e.data);
  // Events: pipeline_started, step_started, step_data_ready, step_completed, pipeline_completed
};

See .env.example for all options. Key settings:

• Airweave Documentation
• Airweave Python SDK

MIT License - use this as a starting point for your own sales context agent.

Built with Airweave - unified search across all your apps and databases.