refactor(routing): rename hint:reasoning-quick → hint:chat#1801
Merged
senamakel merged 1 commit intoMay 15, 2026
Merged
Conversation
Contributor
📝 WalkthroughWalkthroughThis PR introduces a new ChangesReasoning-quick model tier integration
Estimated code review effort🎯 2 (Simple) | ⏱️ ~10 minutes Possibly related PRs
Poem
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
src/openhuman/routing/provider.rs (1)
90-105: 🛠️ Refactor suggestion | 🟠 Major | ⚡ Quick winAdd explicit trace/debug log for remote model normalization.
Line 90-105 adds new
hint:chatnormalization but does not emit a direct normalization log. Add onedebug/traceevent with requested hint/category/resolved model to make routing regressions grepable.🧭 Suggested patch
fn resolve_remote_model(&self, requested_model: &str, category: TaskCategory) -> String { if category != TaskCategory::Heavy { return self.remote_fallback_model.clone(); } - // Keep remote model naming aligned with backend modelRegistry. - match requested_model.strip_prefix("hint:") { + // Keep remote model naming aligned with backend modelRegistry. + let resolved = match requested_model.strip_prefix("hint:") { Some("reasoning") => MODEL_REASONING_V1.to_string(), // Orchestrator's low-TTFT chat tier — Kimi K2.6 Turbo on the // backend's `reasoning-quick-v1`. Backend support added in // tinyhumansai/backend#760. Some("chat") => MODEL_REASONING_QUICK_V1.to_string(), Some("agentic") => MODEL_AGENTIC_V1.to_string(), Some("coding") => MODEL_CODING_V1.to_string(), _ => requested_model.to_string(), - } + }; + tracing::debug!( + requested_model, + category = category.as_str(), + resolved_model = resolved.as_str(), + "[routing] normalized remote model" + ); + resolved }As per coding guidelines, “In Rust, use
log/tracingatdebugortracelevel for development-oriented diagnostics on new/changed flows, including logs at … branch decisions …”.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@src/openhuman/routing/provider.rs` around lines 90 - 105, The resolve_remote_model function currently normalizes hint-prefixed models but lacks an observable log; add a tracing debug/trace call inside resolve_remote_model that logs the input requested_model, the TaskCategory (category), and the final resolved model string (the value returned or remote_fallback_model) so routing changes are grepable — place the log right before each return (or once before the final return) and use the existing tracing/log crate at debug/trace level; reference resolve_remote_model, requested_model, category, remote_fallback_model, and the constants like MODEL_REASONING_QUICK_V1 in the log message.
🧹 Nitpick comments (1)
src/openhuman/providers/router.rs (1)
9-13: ⚡ Quick winUse the canonical model constant in tier mapping.
Line 12 hardcodes
"reasoning-quick-v1"; preferMODEL_REASONING_QUICK_V1to prevent drift between routing layers.♻️ Suggested patch
+use crate::openhuman::config::MODEL_REASONING_QUICK_V1; + fn openhuman_tier_to_hint(model: &str) -> Option<&'static str> { match model { "reasoning-v1" => Some("reasoning"), - "reasoning-quick-v1" => Some("chat"), + MODEL_REASONING_QUICK_V1 => Some("chat"), "agentic-v1" => Some("agentic"), "coding-v1" => Some("coding"), "summarization-v1" => Some("summarization"),🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@src/openhuman/providers/router.rs` around lines 9 - 13, The match in openhuman_tier_to_hint uses a hardcoded "reasoning-quick-v1"; replace that literal with the canonical constant MODEL_REASONING_QUICK_V1 in the match arm inside the openhuman_tier_to_hint function, and ensure the constant is in scope (import or fully qualify it) so the mapping uses the single source-of-truth symbol instead of a string literal.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Outside diff comments:
In `@src/openhuman/routing/provider.rs`:
- Around line 90-105: The resolve_remote_model function currently normalizes
hint-prefixed models but lacks an observable log; add a tracing debug/trace call
inside resolve_remote_model that logs the input requested_model, the
TaskCategory (category), and the final resolved model string (the value returned
or remote_fallback_model) so routing changes are grepable — place the log right
before each return (or once before the final return) and use the existing
tracing/log crate at debug/trace level; reference resolve_remote_model,
requested_model, category, remote_fallback_model, and the constants like
MODEL_REASONING_QUICK_V1 in the log message.
---
Nitpick comments:
In `@src/openhuman/providers/router.rs`:
- Around line 9-13: The match in openhuman_tier_to_hint uses a hardcoded
"reasoning-quick-v1"; replace that literal with the canonical constant
MODEL_REASONING_QUICK_V1 in the match arm inside the openhuman_tier_to_hint
function, and ensure the constant is in scope (import or fully qualify it) so
the mapping uses the single source-of-truth symbol instead of a string literal.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: 4c6b6a54-f946-4e78-94d5-f52e1a657d93
📒 Files selected for processing (8)
src/openhuman/agent/agents/loader.rssrc/openhuman/agent/agents/orchestrator/agent.tomlsrc/openhuman/config/mod.rssrc/openhuman/config/schema/types.rssrc/openhuman/providers/router.rssrc/openhuman/routing/policy.rssrc/openhuman/routing/provider.rssrc/openhuman/routing/provider_tests.rs
…trator Builds on tinyhumansai#1761 (which introduced the `reasoning-quick-v1` tier and wired it up as `hint:reasoning-quick`). Renames the hint slot to `hint:chat` because the orchestrator's actual job — front-line conversational TTFT — is described better by "chat" than by the backend model id leaking into the hint namespace. Same end-to-end behavior as tinyhumansai#1761: hint:chat ↔ reasoning-quick-v1 (Kimi K2.6 Turbo on Fireworks). - `agent/agents/orchestrator/agent.toml`: `hint = "chat"`. - `agent/agents/loader.rs`: test asserts `h == "chat"`. - `providers/router.rs`: `openhuman_tier_to_hint("reasoning-quick-v1")` now returns `Some("chat")` so custom `model_routes` round-trip. - `routing/provider.rs:resolve_remote_model`: `Some("chat") => MODEL_REASONING_QUICK_V1` instead of `Some("reasoning-quick")`. - `routing/policy.rs`: rustdoc documents `hint:chat` as Heavy (always remote — local model can't meet the TTFT budget). - `routing/provider_tests.rs`: new regression `regression_chat_hint_routes_remote_as_reasoning_quick_v1`. - `config/schema/types.rs`: docstring on `MODEL_REASONING_QUICK_V1` references `hint:chat`. Tests: `regression_chat_hint_routes_remote_as_reasoning_quick_v1`, `orchestrator_has_chat_hint_and_named_tools` — both pass.
senamakel
force-pushed
the
fix/hint-chat-translate-to-reasoning-quick-v1
branch
from
2be851c to
b54ee2b
Compare
senamakel
changed the title
Member
Author
|
Note on red CI checks:
This PR only touches |
12 tasks
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
1 participant
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.
Summary
Follow-up to #1761 — renames the orchestrator's hint slot from
reasoning-quicktochat. Same end-to-end behavior; same backend model (reasoning-quick-v1/ Kimi K2.6 Turbo via backend#760).agent/agents/orchestrator/agent.toml:hint = "chat".agent/agents/loader.rs: loader test assertsh == "chat".providers/router.rs:openhuman_tier_to_hint("reasoning-quick-v1") = Some("chat")for custommodel_routesround-trip.routing/provider.rs:resolve_remote_model:Some("chat") => MODEL_REASONING_QUICK_V1.routing/policy.rs: rustdoc documentshint:chatas Heavy (always remote).routing/provider_tests.rs: new regressionregression_chat_hint_routes_remote_as_reasoning_quick_v1.config/schema/types.rs: docstring onMODEL_REASONING_QUICK_V1referenceshint:chat.Problem
hint:reasoning-quick(introduced in #1761) lets the backend model id leak into the hint namespace. The orchestrator's actual job is front-line conversational TTFT, which is named better bychat. Hints should describe the workload, not the backend tier — the same wayhint:reasoningdoesn't sayhint:deepseek-v4-pro.Solution
Rename the single hint slot. Backend mapping (
hint:chat→reasoning-quick-v1) lives entirely client-side inrouting::provider::resolve_remote_model, exactly where the other hint-to-model translations live. No backend changes needed — the constantMODEL_REASONING_QUICK_V1and the backend model id are unchanged.Submission Checklist
diff-cover) meet the gate enforced by.github/workflows/coverage.yml.docs/TEST-COVERAGE-MATRIX.mdreflect this change## Relateddocs/RELEASE-MANUAL-SMOKE.md)Closes #NNNin the## RelatedsectionImpact
reasoning-quick-v1on the backend. Only the hint string inagent.tomland the route-table key change.model_routesneed achatentry instead ofreasoning-quick. Without one, the request falls through to the default provider model (same fallback as any other unmapped hint).Related
AI Authored PR Metadata (required for Codex/Linear PRs)
Linear Issue
Commit & Branch
Validation Run
Validation Blocked
Behavior Changes
Parity Contract
reasoning-quick-v1) and constant (MODEL_REASONING_QUICK_V1) are unchanged. Cost rows and identity-cost defaults from feat(orchestrator): use reasoning-quick-v1 for low-latency chat #1761 stay put.routing::policy::classifystill bucketshint:chatasHeavy;routing::provider::resolve_remote_modeltranslateshint:chatto the same backend model id feat(orchestrator): use reasoning-quick-v1 for low-latency chat #1761 used. Regression test confirms the end-to-end remote-model string.Duplicate / Superseded PR Handling