This commit implements two main changes based on your feedback:
1. Refactors all Model Context Protocol (MCP) resources into tools:
- In `whatsappIntegration.js`:
- `whatsapp.chat` (resource) -> `whatsapp.list-chats` (tool) & `whatsapp.get-chat-details` (tool)
- `whatsapp.contact` (resource) -> `whatsapp.get-contact-details` (tool)
- `whatsapp.blocklist` (resource) -> `whatsapp.get-blocklist` (tool)
- In `conversationIntegration.js`:
- `conversation` (resource) -> `conversation.list-conversations` (tool) & `conversation.get-conversation-details` (tool)
This change aligns with current Gemini SDK capabilities for MCP.
2. Implements chat deletion functionality:
- In `whatsapp-router`:
- Added `deleteChat` function to `whatsappClient.ts`. This function calls an assumed OpenWA endpoint (`POST /deleteChat`) and will require verification against the actual OpenWA API.
- Added a `DELETE /chats/:chatId` route to `whatsappActions.ts` that utilizes the new `deleteChat` client function.
- In `conversation-layer-mcp`:
- Added a new `whatsapp.delete-chat` tool to `whatsappIntegration.js`. This tool calls the new `DELETE /chats/:chatId` endpoint in `whatsapp-router`.
Additionally, the existing `conversation.delete` MCP tool was verified to be correctly implemented.
This commit introduces a significant refactoring to the whatsapp-router service to improve modularity and extend its capabilities for interacting with OpenWA (nucleo-whatsapp).
Key changes include:
1. **Message Processing Logic**:
* Created a new `messageProcessor.ts` module.
* Moved message mapping (`mapWhatsAppMessageToMsg`), message addition to conversations (`processIncomingMessageForConversation`), and audio transcription handling (`handleAudioMessageTranscription`) into this new module.
* Updated `store/conversation.ts` and `webhook.ts` to utilize `messageProcessor.ts`, streamlining their responsibilities.
2. **OpenWA Client (`nucleoClient.ts`)**:
* Introduced `nucleoClient.ts`, a dedicated module for encapsulating API calls to OpenWA.
* Implemented functions for various OpenWA commands:
* Sending text, image, and file messages.
* Fetching chat and contact information.
* Creating groups.
* Managing the blocklist (get, block, unblock).
* Listing all chats and fetching messages for a specific chat.
* Includes error handling and basic typing for API responses.
3. **New API Endpoints (`routes/nucleoActions.ts`)**:
* Created `nucleoActions.ts` to expose the functionalities of `nucleoClient.ts` via a new set of HTTP API endpoints.
* Endpoints cover all implemented client functions, with request validation and robust error handling.
* These routes are grouped under the `/nucleo` base path.
4. **Route Registration**:
* Registered the new `/nucleo` routes in the main `index.ts` file.
5. **Conversation Store and Routes Review**:
* Reviewed and confirmed that `store/conversation.ts` and `routes/conversations.ts` are correctly integrated with the new `messageProcessor.ts` and remain focused on their core responsibilities.
This refactoring enhances the structure of `whatsapp-router`, making it easier to maintain and extend. It also provides a comprehensive set of API endpoints for more granular control over OpenWA functionalities.
This commit modifies the audio download process in `whatsapp-router/src/webhook.ts` to align with the updated nucleo-whatsapp API.
Previously, audio files were downloaded using a direct GET request to the file URL. This has been changed to a POST request to the `/downloadFileWithCredentials` endpoint provided by nucleo-whatsapp.
Key changes:
- Audio files are now downloaded by POSTing to `[OPEN_WA_URL]/downloadFileWithCredentials` with the audio file's URL in the request body.
- The `/downloadFileWithCredentials` endpoint returns a base64 encoded string directly, so the explicit base64 conversion step after downloading has been removed.
This change ensures compatibility with the correct API for fetching message media.
This commit removes the Jest unit tests, configuration, and related dependencies that were added for the WhatsApp audio transcription feature.
This is done as per your request due to potential issues with the test execution environment.
The core functionality for audio transcription remains.
