Get embedding from cache.

Run the CLI main loop.

Log a debug message if the DEBUG flag is set to True

Run the CLI main loop.

Returns the parameter value. If the parameter value is a file path, it will be read and returned.

Returns the parameter value. If the parameter value is a file path, it will be read and returned.

Perform pure semantic search using vector similarity. Args: ctx: Agent runtime context with dependencies query: Search q

Returns a default resultset

Get embedding from cache.

Return an error resultset.

Search the knowledge base for relevant information. Args: params: Search parameters Returns

Set a user preference for the session.

Get embedding from cache.

Save session data for a Telegram user. Args: user_id: Telegram user ID session_id: Claude SDK session ID cwd: Workin

Use fine-tuned model for embeddings

Prints the message for debugging

Log a message to the Supabase 'messages' table. Args: session_id (str): Unique session identifier. message_type (str): 'huma

Acquire a connection from the pool.

Perform hybrid search combining semantic and keyword matching. Args: ctx: Agent runtime context with dependencies query:

Acquire a connection from the pool.

Initialize external connections.

Acquire a connection from the pool.

Close connection pool.

Perform a LLM refined prompt request and messages array generation It's a wrapper for the get_messages_array and get_refined_prompt

Generate embedding for text using OpenAI.

Load session data for a Telegram user. Args: user_id: Telegram user ID Returns: Tuple of (session_id, cwd) if session e

Acquire a connection from the pool.

Add a query to the search history.

Clean up external connections.

Establishes connection to MCP server

Load settings with proper error handling.

Show a form submission error

Chunk document using simple rules. Args: content: Document content title: Document title

Run the CLI main loop.

Starts each MCP server and returns the tools for each server formatted for Pydantic AI.

Returns the model parameters for the LLM call Args: messages (list): The messages array for the LLM call.

Get the working directory for a user. Args: user_id: Telegram user ID Returns: Working directory path (falls back to WO

Perform a X AI (Grok) request

Create embedding generator with optional caching. Args: model: Embedding model to use use_cache: Whether to use caching

Generate embedding for a search query. Args: query: Search query Returns: Query emb

Initialize Graphiti client.

Returns the current time and prints the message

Send a query to the Claude SDK. Args: user_prompt: The user's question or request Note: This initia

Create appropriate chunker based on configuration. Args: config: Chunking configuration Returns: Chunker instan

Returns the LLM text model

Initialize the JSON file database

Set the working directory for a user. Args: user_id: Telegram user ID cwd: Working directory path

Create a callable function for a specific tool. This allows us to execute functions through the MCP server. Args:

Clear the session for a user (keeps cwd configuration). Args: user_id: Telegram user ID

Returns the LLM provider

Returns the LLM provider

Returns the OpenAI API response for a LLM request

Handle regular text messages and interface with Claude SDK. This function: 1. Loads user's session and cwd configuration 2. Configur

Initialize database connection pool.

Save the conversation in the database

Search the knowledge graph. Args: query: Search query center_node_distance: Distance from center nod

Send a long message, splitting it if necessary due to Telegram's 4096 character limit. Args: chat_id: Telegram chat ID text:

Add an episode to the knowledge graph. Args: episode_id: Unique episode identifier content: Episode

Add a message to a session. Args: session_id: Session UUID role: Message role (user/assistant/system) content: M

Split text into chunks using a simple sliding window approach. Args: text: The text to split into chunks

Extract text from a TXT file with encoding fallbacks. Args: file_path: Path to the TXT file

Returns the app configuration

Returns the available LLM providers based on the environment variables The model will be available if all its variables are set

Get the appropriate processor for a file based on its extension. Args: file_path: Path to the document file Returns

Returns the LLM text model

Initialize the JSON file database

Initialize the server connection.

Receive messages until and including ResultMessage. This method waits for the complete response and returns all messages. Us

Store a message in the Supabase messages table.

Stores a message in the Supabase `MESSAGES` table. This function inserts a new message into the `MESSAGES` table, including the session ID, m

Validate the question

Simple text splitting as fallback. Args: text: Text to split Returns: List of chunk

Close connection pool.

Close connection pool.

Create embedding generator with optional caching. Args: model: Embedding model to use use_cache: Whether to use caching

Generate embedding for a search query. Args: query: Search query Returns: Query emb

Get embedding from cache.

Get the result of an ideation request based on the provided text and button index. Args: user_query: The text to base the ideati

Returns the LLM model

Returns the LLM model

Returns the model options for the LLM call

Get the new unique item id

Returns the parameter value or the environment variable value

Fetch season averages for multiple players.

Get session by ID. Args: session_id: Session UUID Returns: Session data or None if not found/expired

Fetch current injuries for a team

Reads a file and returns its content. If the file_path is a URL, it will be downloaded If the file_path is a local file, it will be read

Recycle the suggestions from the AI

Reset all state and refresh the page

Show buttons based on the buttons_config Args: buttons_config (listo): list of buttons configurations [

Store a message in the Supabase messages table.