fix: keep agents on the latest request after tool search

[senamakel]

Summary

• Pin TinyAgents with the fix for prompt-guided tool continuations after a search result.
• Keep the latest user request at the end of the outgoing model prompt so a later tool result does not pull the model back to an earlier greeting.

Problem

The supplied OpenHuman transcript shows “fetch my latest email” and a later Composio request. tool_search found GMAIL_FETCH_EMAILS and returned its schema, but DeepSeek V4 Flash answered “Hey! What's up?” after the tool result. The model-call journal contains the full email request and search result; the response's reasoning and text refer to the initial greeting.

Solution

The fix lives in tinyinference and TinyAgents, which own prompt-guided message conversion and the agent loop. This PR pins the updated TinyAgents commit. The outgoing request repeats the latest real text user request immediately after a terminal tool-result turn. The durable transcript is unchanged.

tinyhumansai/tinyinference#29 and tinyhumansai/tinyagents#216 have merged. The pinned commits are available upstream; this PR is ready for review.

Submission Checklist

• Tests added or updated: upstream regressions cover the greeting/email search and a history without a real user request.
• Diff coverage ≥ 80% — N/A here: this PR changes only a gitlink; upstream CI owns changed-line coverage. CI pending.
• Coverage matrix updated — N/A: no OpenHuman feature row changed.
• All affected feature IDs listed under Related — N/A: no matrix feature changed.
• No new external network dependencies introduced.
• Manual smoke checklist updated — N/A: no release-cut surface changed.
• Linked issue closed — N/A: no linked issue.

Impact

Desktop, CLI, TUI, and library agents using a text tool dialect can remain focused on the latest request after tool results. One extra user-role reminder is sent to the provider on those continuations, increasing prompt tokens by the length of the current text request. The persisted transcript and tool schema are unchanged.

Related

• Closes: N/A
• Follow-up PR(s)/TODOs: fix: keep active request after prompt-guided tool results tinyinference#29, fix: ground text tool continuations in the latest request tinyagents#216

---

AI Authored PR Metadata (required for Codex/Linear PRs)

Linear Issue

• Key: N/A
• URL: N/A

Commit & Branch

• Branch: transcript-hi-investigation
• Commit SHA: c86275e

Validation Run

• pnpm --filter openhuman-app format:check — N/A: no app changes
• pnpm typecheck — N/A: no TypeScript changes
• Focused tests: tinyinference prompt_tools 17 passed; TinyAgents dialect 6 passed
• Rust fmt/check (if changed): both upstream Rust workspaces formatted; OpenHuman cargo check -p openhuman passed
• Tauri fmt/check (if changed) — N/A: no Tauri changes

Validation Blocked

• command: N/A
• error: N/A
• impact: N/A: upstream dependencies merged

Behavior Changes

• Intended behavior change: keep the current user request after a text-dialect tool result in the outgoing model prompt.
• User-visible effect: the agent should continue the requested email lookup rather than answer the earlier greeting.

Parity Contract

• Legacy behavior preserved: durable transcript and native tool calls stay unchanged.
• Guard/fallback/dispatch parity checks: upstream tests cover text-dialect dispatch and a result-only history.

Duplicate / Superseded PR Handling

• Duplicate PR(s): N/A
• Canonical PR: this PR
• Resolution (closed/superseded/updated): N/A

Summary by CodeRabbit

• Updates
  • Updated a bundled component to a newer version. The available change details do not specify any changes to user-facing features or behavior, so no further impact can be confirmed. No new capabilities, removals, or fixes are documented for this update.

[coderabbitai]

Navigate logical layers of code changes, visualize relationships, and explore their blast radius.

Note

Currently processing new changes in this PR. This may take a few minutes, please wait...

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Advanced

Run ID: 2b881da1-ebcb-4003-8141-b3b18c6fef0c

📥 Commits

Reviewing files that changed from the base of the PR and between 573d1d4 and c86275e.

📒 Files selected for processing (1)

• vendor/tinyagents

 __________________________________________________________________________________________________
< Estimate to avoid surprises. Estimate before you start. You'll spot potential problems up front. >
 --------------------------------------------------------------------------------------------------
  \
   \   \
        \ /\
        ( )
      .( o ).

---

_{Comment @coderabbitai help to get the list of available commands.}

[tinysweeper]

Tiny Sweeper review

Tiny Sweeper reviewed this change across 6 lane(s) and found 0 active actionable finding(s). Detailed lane evidence and any incomplete work are listed below.

State: Ready for maintainer review
Priority: none
Reviewed head: c86275e3671f
Updated: 1790334659 (Unix time)

Review snapshot

Change surface	Files	Review signal	Count
Production	0	Active findings	0
Tests	0	Noted findings	0
Documentation	0	Resolved findings	0
Configuration	0	Pending checks/questions	0

Completeness: Complete
Test assessment: No supported feature-to-test mapping was available; this does not mean tests are absent or passed.

What changed

The review could not produce a supported behavioral summary; inspect the cited changed surface and lane details below.

Features

None identified with supported citations.

Tests

No supported feature-to-test mapping was produced. Test execution is not inferred.

Findings

No active actionable findings.

Before merge

None.

Agent review details

critique

• Conclusion: Neutral
• Scope reviewed: all assigned evidence
• Lane summary: No added or modified lines to review.

security

• Conclusion: Neutral
• Scope reviewed: all assigned evidence
• Lane summary: No added or modified lines to review.

tests

• Conclusion: Neutral
• Scope reviewed: all assigned evidence
• Lane summary: No added or modified lines to review.

commits

• Conclusion: Neutral
• Scope reviewed: all assigned evidence
• Lane summary: Nothing sensitive found in what this pull request commits.

description

• Conclusion: Neutral
• Scope reviewed: all assigned evidence
• Lane summary: No added or modified lines to review.

e2e

• Conclusion: Neutral
• Scope reviewed: all assigned evidence
• Lane summary: No added or modified lines to review.

Evidence and run details

• Models: ladder/vectors
• Spend: $0.000000
• Tokens: 0 input · 0 output · 0 cached · 13 embedding

Head	State	Pass summary
c86275e3671f	ready for maintainer review	0 active finding(s), 0 resolved finding(s) (at 1790334659)

[tinysweeper]

tinysweeper found nothing blocking. Approving.

$0.0000 · 0 in / 0 out · 13 embedded · ladder/vectors