AI Letter Generation Agent

Built with Langchain

This project is an AI-powered automation tool for generating professional company letters and exporting them to Google Docs. It features a FastAPI backend, a modern chat-based frontend, and deep integration with Google Drive and OpenAI (or compatible LLMs) via LangChain.

---

📁 File Structure

report_agent copy/
├── backend/
│   ├── app.py                        # FastAPI backend entrypoint and API endpoints
│   ├── letter_agent.py               # Core automation logic for letter generation (LangChain agent)
│   ├── google_docs.py                # Google Docs API integration for document creation
│   ├── client_secret.json            # [PLACEHOLDER] Google OAuth client credentials (do not commit real secrets)
│   ├── report-agent-...json          # [PLACEHOLDER] Google service account credentials (do not commit real secrets)
│          
├── frontend/
│   └── main.html                     # Modern chat UI for interacting with the agent
└──              

---

🚀 Setup & Installation

1. Clone the repository

2. Install dependencies

   pip install -r requirements.txt

3. Configure credentials

   • Place your Google OAuth and service account credentials in backend/client_secret.json and backend/report-agent-...json (see the provided placeholder files for structure).
   • Set your OpenAI or compatible API key as an environment variable (OPENAI_API_KEY).
   • Update any other placeholders in letter_agent.py and google_docs.py (e.g., Google Drive folder IDs, template IDs).

4. Run the backend

   uvicorn backend.app:app --reload

5. Open the frontend

   • Open frontend/main.html in your browser.
   • Ensure the backend is running at http://localhost:8000 (or update the API URL in the HTML if needed).

---

🧠 Workflow: How Letter Automation Works

1. User Interaction

• The user interacts with a chat interface (frontend/main.html), sending prompts describing the letter they want to generate.

2. API Request

• The frontend sends the user's message to the /chat endpoint of the FastAPI backend (backend/app.py).

3. LetterGenerationAgent (Core Logic)

• The backend manages conversation state and delegates the request to a LetterGenerationAgent (see backend/letter_agent.py).
• The agent uses LangChain's ReAct framework, combining:
  • Company context: Loads relevant documents from Google Drive (via GoogleDriveLoader).
  • LLM reasoning: Uses OpenAI (or compatible) LLM to generate letter content, recipient details, and role.
  • Custom tools:
    • get_current_datetime: Fetches the current date/time.
    • get_company_context: Extracts company/recipient info from loaded docs.
    • generate_formal_letter: Assembles a formal letter using a template and LLM-generated content.
    • create_google_doc_from_letter: Converts the letter into a Google Doc and returns the shareable URL.
• The agent follows a workflow:
  1. Extract recipient info from company docs.
  2. Generate main content and role using the LLM.
  3. Assemble the letter in a formal template.
  4. Optionally, create a Google Doc and return the link.

4. Response

• The backend returns the generated letter (and optionally a Google Doc link) to the frontend, which displays it in the chat.

---

📝 API Endpoints

• GET / — Health check
• POST /chat — Main chat endpoint (send user prompt, get letter response)
• POST /new-conversation — Start a new chat session
• DELETE /conversation/{conversation_id} — Delete a session
• GET /conversations — List all active sessions

---

⚙️ Dependencies

See requirements.txt for all dependencies. Key packages:

• fastapi, uvicorn — API server
• langchain, langchain-openai, langchain-google-community — LLM and tool orchestration
• openai — LLM API
• google-auth, google-api-python-client, google-auth-oauthlib — Google Docs/Drive integration
• python-dotenv — Environment variable management

---

Built by Shilok Kumar