Add scFM execution workflow and route to single_cell_team#97
Open
HendricksJudy wants to merge 3 commits intoaristoteleo:mainfrom
Open
Add scFM execution workflow and route to single_cell_team#97HendricksJudy wants to merge 3 commits intoaristoteleo:mainfrom
HendricksJudy wants to merge 3 commits intoaristoteleo:mainfrom
Conversation
Three GUI usability issues traced earlier are still present and addressed together so a "run scFM on my data" request can actually complete: 1. Two overlapping routers — hide the deprecated `scfm_router` tool from LLMs via `@tool(exclude=True)` and stop appending a literal "DEPRECATED: ..." string to its `warnings[]`, which the chatroom was rendering back to users. The method stays callable from Python for tests and internal helpers; `route_query` is unchanged. 2. The router plan had no executor — add `scfm` to `analysis_expert`'s toolsets and document a short scFM execution workflow telling the agent to dispatch each step from `fm_router`'s plan as a typed `scfm_*` tool call instead of re-implementing it in raw Python. 3. The default General Team had no path to scFM work — add a short "scFM tasks" section to the default leader prompt that, on a clear scFM signal (model name or capability), tells the user once to switch the chat to `single_cell_team` rather than attempting the work without the right tools. Bump default template to 1.5.0.
…cess Address Codex review on PR #3: the scFM execution workflow conflated all non-"ready" outcomes from scfm_preprocess_validate. Split into branches so the agent stops cleanly on "incompatible" (where scfm_run can only echo the same error) and only proceeds through scfm_interpret_results after scfm_run actually succeeds.
…ability-SZfkS Add scFM execution workflow and route to single_cell_team
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This PR integrates single-cell foundation model (scFM) task execution into the analysis expert agent and establishes proper routing for scFM requests. It adds comprehensive execution guidance for the scfm toolset, deprecates the old scfm_router tool in favor of a template-based router, and updates the default team to redirect scFM requests to the dedicated single_cell_team.
Key Changes
analysis_expert.md: Added scfm to the toolsets list and documented a complete scFM execution workflow section that:
Explains how to execute routing decisions from fm_router as typed tool calls
Specifies the typical execution order (preprocess → run → interpret)
Describes handling of validation results and auto-fixes
Clarifies when to fall back to raw Python for custom work
default.md:
Bumped team version from 1.4.0 to 1.5.0
Added "Single-Cell Foundation Model (scFM) tasks" section that instructs the leader agent to redirect scFM requests to single_cell_team instead of attempting them locally
Provides clear guidance on recognising scFM signals (by model name or capability)
toolset.py:
Marked scfm_router tool with exclude=True to hide it from LLM agents
Updated docstring to clarify deprecation and redirect users to the template-based fm_router sub-agent
Removed the in-band deprecation warning from the result dictionary (now only in docstring)
Implementation Details
The changes establish a clear separation of concerns: the default team recognises scFM requests and routes them to single_cell_team, which uses fm_router for model selection and analysis_expert for execution. The old scfm_router tool remains available for internal/test use but is hidden from LLM agents to prevent confusion.