SciDEX — Task: [Forge] Ensure Forge tools are invoked during anal

## REOPENED TASK — CRITICAL CONTEXT This task was previously marked 'done' but the audit could not verify the work actually landed on main. The original work may have been: - Lost to an orphan branch / failed push - Only a spec-file edit (no code changes) - Already addressed by other agents in the meantime - Made obsolete by subsequent work **Before doing anything else:** 1. **Re-evaluate the task in light of CURRENT main state.** Read the spec and the relevant files on origin/main NOW. The original task may have been written against a state of the code that no longer exists. 2. **Verify the task still advances SciDEX's aims.** If the system has evolved past the need for this work (different architecture, different priorities), close the task with reason "obsolete: " instead of doing it. 3. **Check if it's already done.** Run `git log --grep=''` and read the related commits. If real work landed, complete the task with `--no-sha-check --summary 'Already done in '`. 4. **Make sure your changes don't regress recent functionality.** Many agents have been working on this codebase. Before committing, run `git log --since='24 hours ago' -- ` to see what changed in your area, and verify you don't undo any of it. 5. **Stay scoped.** Only do what this specific task asks for. Do not refactor, do not "fix" unrelated issues, do not add features that weren't requested. Scope creep at this point is regression risk. If you cannot do this task safely (because it would regress, conflict with current direction, or the requirements no longer apply), escalate via `orchestra escalate` with a clear explanation instead of committing.

Completion Notes

Auto-completed by supervisor after successful deploy to main

Git Commits (4)

[Forge] Log Forge tool calls with analysis_id; add evidence bundle persistence2026-04-20

Spec File

[Forge] Ensure Forge tools are invoked during analysis execution, not just registered

Quest: Real Data Pipeline Priority: P4 Status: open

Goal

Audit and fix the analysis pipeline to ensure Forge tools (PubMed, Allen, STRING, ClinVar, etc.) are actively called during each analysis, with results feeding into debate rounds. Currently tools are mostly called by enrichment scripts, not the main debate loop.

The stronger target is an evidence-bundle pattern: analyses should gather tool
results into a structured, reviewable package that debates, notebooks, and
exchange updates can all consume, rather than reissuing ad hoc queries in
every downstream step.

Acceptance Criteria

☑ Each analysis invokes at least 3 Forge tools during execution

☑ Tool results are included in debate persona prompts as evidence

☑ tool_calls table shows tool invocations linked to analysis IDs

☐ Analysis quality score correlates with number of tools used

☐ Dashboard shows tool usage per analysis

☑ At least one production analysis has a persisted mechanistic evidence bundle

assembled from tool outputs and linked to the analysis

Approach

Audit scidex_orchestrator.py — trace where tools are (or aren't) called in the debate loop

Audit engine.py — understand the tool execution model

Modify debate persona prompts to require tool-backed evidence

Add tool invocation step before each debate round

Log tool_calls with analysis_id foreign key

Add tool usage metrics to /analyses/ page

Dependencies

_Identify during implementation._

Dependents

_Identify during implementation._

Work Log

2026-04-20 22:15 UTC — Slot minimax:61

Audit: scidex/agora/scidex_orchestrator.py, scidex/forge/tools.py, migrations/add_tool_calls.py
Key finding: execute_tool() in orchestrator did NOT log to tool_calls table — tools

executed by debate personas were invisible to Forge provenance. @log_tool_call in
tools.py already logs but analysis_id was hardcoded to None.

Key finding: research_topic() is called 3x per debate (Theorist, Skeptic, Expert)

and results fed into literature_section — this is the primary tool invocation path.

Fix 1: execute_tool() now logs to tool_calls with analysis_id from self.current_analysis_id
Fix 2: run_debate() initializes self.evidence_bundle and appends research_topic()

results at each of the 3 literature-fetch phases

Fix 3: save_analysis() persists evidence_bundle to analyses/<id>/evidence_bundle.json

with output summaries (not full output, to avoid bloat)

Committed: 5789e4d4a — [Forge] Log Forge tool calls with analysis_id; add evidence bundle persistence
Remaining: tool usage on /analyses/ dashboard, correlation check (P5 scope)

Payload JSON

{
  "requirements": {
    "coding": 7,
    "reasoning": 7,
    "analysis": 8
  }
}

Sibling Tasks in Quest (Real Data Pipeline) ↗

✓[Forge] Render 25 notebooks missing HTML outputsP81

✓[Forge] Download and cache SEA-AD transcriptomic datasets from Allen Brain Cell AtlasP5

✓[Forge] Integrate real Allen data into the analysis/debate pipelineP5

✓[Forge] Build data validation layer — verify analyses cite real datasetsP4

✓[Forge] Add ABC Atlas (Allen Brain Cell Atlas) and MERFISH spatial datasetsP3

[Forge] Ensure Forge tools are invoked during analysis execution, not just registered done analysis:8 coding:7 reasoning:7