tools.cross_channel_query module

Cross-channel query tool – privately ask the LLM about another channel.

Fetches recent messages from a target channel’s Redis cache, builds a one-shot LLM sub-call with that transcript as context, and returns the response. Nothing is posted to the target channel.

async tools.cross_channel_query.run(channel_id='', question='', message_count=100, *, ctx=None)[source]

Public tool entry point that delegates to _query_channel().

Thin wrapper exposing the cross-channel query under the loader’s expected run signature: it forwards every argument straight to the private _query_channel() implementation, which does the real work (fetching the target channel’s transcript from the Redis message cache or platform API, optionally folding in the cached summary, and running a tool-less one-shot LLM sub-call). All side effects live in that callee.

Called by the tool dispatcher: tool_loader.py resolves this module’s run via getattr(module, "run") and registers it under TOOL_NAME (“query_channel”); no direct in-repo callers were found.

Parameters:

  • channel_id (str) – Target channel/room ID to read from.

  • question (str) – The question to answer about that channel’s activity.

  • message_count (int) – Number of recent messages to include as context (default 100, capped at 500 inside the callee).

  • ctx (ToolContext | None) – Injected tool context (Redis, adapter, config, message cache).

Returns:

The JSON result string produced by _query_channel().

Return type:

str