A powerful AI assistant integration for Pinecone Assistant.
-
Install dependencies:
npm install
-
Test the integration:
npm test
- Authentication: API key-based authentication with Pinecone Assistant
- Create Assistant: Create new Pinecone Assistants with custom instructions
- Update Assistant: Modify existing assistant instructions and metadata
- Delete Assistant: Remove existing assistants
- Upload Document: Upload documents to assistants for context-aware responses
- List Assistants: Search and list all available assistants in your project
- Check Assistant Status: Get detailed status and information for a specific assistant
- List Files: List all files uploaded to a specific assistant
- Describe File: Get detailed status and metadata of a specific file
- Chat with Assistant: Chat with assistants and get citations in structured form
- Retrieve Context: Retrieve context snippets for RAG or agentic workflows
- Evaluate Answer: Evaluate correctness and completeness of assistant responses
- New Assistant Trigger: Triggers when a new assistant is created
- New File Trigger: Triggers when a new file is uploaded to an assistant
- File Status Change Trigger: Triggers when a file's processing status changes
You'll need a Pinecone API key to use this integration. You can find this in your Pinecone dashboard.
This integration follows the standard app structure:
index.js
- Main app configurationauthentication.js
- Authentication configurationcreates/
- Create operations (Assistant and Document)searches/
- Search operations (List Assistants, Chat, etc.)triggers/
- Trigger operations (New Assistant)test/
- Test files
Creates a new Pinecone Assistant with:
- Name (required)
- Custom instructions (optional)
Updates an existing assistant with:
- Assistant Name (required)
- Instructions (required) - New instructions for the assistant
- Metadata (optional) - JSON string of metadata to update
Deletes an existing assistant with:
- Assistant Name (required)
Uploads documents to an existing assistant with:
- Assistant ID (required)
- Document file (required)
- Optional metadata
Lists all available assistants in your project with:
- No input fields required
- Returns assistant details including name, instructions, status, and timestamps
Gets detailed status and information for a specific assistant with:
- Assistant Name (required)
- Returns complete assistant details including status, instructions, metadata, host, and timestamps
Lists all files uploaded to a specific assistant with:
- Assistant Name (required)
- Metadata Filter (optional) - JSON-encoded metadata filter for files
- Returns file details including name, ID, metadata, status, processing progress, and timestamps
Gets detailed status and metadata of a specific file with:
- Assistant Name (required)
- File ID (required) - The UUID of the file to describe
- Include Signed URL (optional) - Whether to include the signed URL in the response
- Returns complete file details including status, processing progress, download URL, and error messages
Chat with an assistant and get citations in structured form with:
- Assistant Name (required)
- Message (required) - The message to send to the assistant
- Model (optional) - LLM model to use (GPT-4o, Claude, Gemini, etc.)
- Temperature (optional) - Controls randomness (0 = deterministic)
- Stream Response (optional) - Whether to stream responses
- JSON Response (optional) - Whether to return JSON format
- Include Highlights (optional) - Whether to include document highlights
- Document Filter (optional) - JSON filter to limit document retrieval
- Returns assistant response with citations, usage stats, and referenced documents
Retrieve context snippets from an assistant for RAG or agentic workflows with:
- Assistant Name (required)
- Query (optional) - The query to generate context from
- Messages (optional) - JSON array of messages (alternative to query)
- Top K (optional) - Maximum snippets to return (default: 16, max: 64)
- Snippet Size (optional) - Maximum snippet size in tokens (default: 2048)
- Document Filter (optional) - JSON filter to limit document retrieval
- Returns context snippets with scores, file references, and page numbers
Evaluate the correctness and completeness of assistant responses with:
- Question (required) - The question for which the answer was generated
- Generated Answer (required) - The answer from the assistant or RAG system
- Ground Truth Answer (required) - The correct answer for comparison
- Returns evaluation metrics including correctness, completeness, alignment scores, and detailed reasoning
Triggers when a new assistant is created with:
- No input fields required
- Returns assistant details including name, instructions, status, host, and timestamps
- Polling trigger that checks for new assistants periodically
Triggers when a new file is uploaded to an assistant with:
- Assistant Name (required) - The name of the assistant to monitor for new files
- Returns file details including name, ID, metadata, status, processing progress, and timestamps
- Polling trigger that checks for new files in the specified assistant
Triggers when a file's processing status changes with:
- Assistant Name (required) - The name of the assistant to monitor for file status changes
- Status Filter (optional) - Only trigger for specific status changes (Available, Failed, Processing, Deleting)
- Returns file details including name, ID, metadata, status, processing progress, and timestamps
- Polling trigger that monitors file processing status changes