Refactor inference around external Ollama routing#1975
Merged
senamakel merged 22 commits intoMay 17, 2026
Merged
Conversation
Contributor
|
Note Reviews pausedIt looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the Use the following commands to manage reviews:
Use the checkboxes below for quick actions:
📝 WalkthroughWalkthroughAdds a new inference namespace, routes AI RPCs through it, treats Ollama as an externally managed runtime (removes in-app install/download/path controls), updates provider APIs (temperature suppression, model listing), implements OpenAI-compatible /v1 HTTP endpoints, and updates frontend/backend tests and CI e2e tooling. ChangesInference externalization and routing
Sequence Diagram(s)sequenceDiagram
participant App as Frontend
participant RPC as Tauri / frontend RPC
participant Core as Core inference RPC
participant Local as Local inference service
participant Ext as External runtime (Ollama / LM Studio)
App->>RPC: inference_prompt
RPC->>Core: openhuman.inference_prompt
Core->>Local: inference_prompt(...)
alt provider = ollama/lm_studio
Local->>Ext: HTTP /v1/chat or Ollama API
Ext-->>Local: response/stream
else cloud provider
Local->>Ext: provider endpoint (with auth)
Ext-->>Local: response/errors
end
Local-->>Core: RpcOutcome
Core-->>RPC: result
RPC-->>App: render status/guidance
Estimated code review effort🎯 5 (Critical) | ⏱️ ~120 minutes Possibly related PRs
|
…ernal-ollama-routing
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Actionable comments posted: 4
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (2)
src/openhuman/local_ai/README.md (1)
3-3:⚠️ Potential issue | 🟡 Minor | ⚡ Quick winUpdate the module synopsis to match external-runtime ownership.
Line 3 still says local_ai “owns the bundled Ollama runtime” and “download + install management,” which contradicts the new external-runtime contract.
🤖 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/local_ai/README.md` at line 3, Update the module synopsis in the README for the local_ai module to remove claims that it "owns the bundled Ollama runtime" and "download + install management" and instead state that external-runtime now owns the bundled runtimes and asset install/management; keep references to what local_ai still owns (e.g., LM Studio integration, whisper.cpp, Piper, sentiment scoring, vision-embedding routing, GIF heuristic, and the per-session LocalAiService singleton) and clarify that remote-provider HTTP transport (providers/) and agent tool loop (agent/) remain out of scope.src/openhuman/local_ai/service/ollama_admin.rs (1)
46-52:⚠️ Potential issue | 🟡 Minor | ⚡ Quick winFix stale contract docs for
ensure_ollama_server_fresh.The comment says this function forces a fresh install, but Line 52 now just delegates to
ensure_ollama_serverwith no install/start behavior.Suggested patch
- /// Like `ensure_ollama_server`, but forces a fresh install of the Ollama binary - /// (ignoring cached/workspace binaries). Used as a retry after the first attempt fails. + /// Retry wrapper around `ensure_ollama_server`. + /// Kept for call-site compatibility; does not perform install/start side effects. pub(in crate::openhuman::local_ai::service) async fn ensure_ollama_server_fresh( &self, config: &Config, ) -> Result<(), String> { self.ensure_ollama_server(config).await }🤖 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/local_ai/service/ollama_admin.rs` around lines 46 - 52, The doc comment for ensure_ollama_server_fresh is stale: the function currently just delegates to ensure_ollama_server but the comment claims it forces a fresh install; either update the comment to reflect delegation or implement the fresh-install behavior. To fix, either (A) change the doc for ensure_ollama_server_fresh to state that it currently calls ensure_ollama_server and does not force reinstall, or (B) implement the intended behavior by making ensure_ollama_server_fresh clear any cached/workspace binaries (or call an existing install helper with a "force" flag) and then call the install/start routines before returning; locate ensure_ollama_server_fresh and ensure_ollama_server to add the cache-clear or force-install logic or to update the comment accordingly.
🧹 Nitpick comments (5)
src/openhuman/local_ai/service/ollama_admin_tests.rs (1)
126-219: 🏗️ Heavy liftSplit this test module before adding more cases.
This file is already far above the Rust module size target, and these additions increase the maintenance/merge-conflict surface. Please move the new external-runtime diagnostics tests into a dedicated sibling test module.
As per coding guidelines:
src/**/*.rs: Keep Rust source files to ≤ ~500 lines; split modules when growing larger.🤖 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/local_ai/service/ollama_admin_tests.rs` around lines 126 - 219, The test module has grown too large; move the three new tests (ensure_ollama_server_requires_external_runtime_when_unreachable, ensure_ollama_server_reports_broken_external_runner_without_restart_attempt, assets_status_marks_ollama_unavailable_when_runtime_is_down_even_if_binary_exists) into a new sibling test module file to reduce module size and merge-conflict surface. Create a separate test-only module file, import the same helpers (e.g., local_ai_test_guard, spawn_mock, Config, LocalAiService) and any env setup/teardown logic, copy those three test functions verbatim into it, and remove them from the original file; ensure the new file compiles by adding the necessary use statements and keeping the #[tokio::test] annotations intact.app/src/components/settings/panels/local-model/DeviceCapabilitySection.tsx (1)
45-51: ⚡ Quick winPrune legacy install-state contract from this component.
These constants make install/retry paths permanently unreachable while the related props remain in the API. Remove the obsolete install props/branches now to avoid dead UI paths and no-op parent wiring.
🤖 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 `@app/src/components/settings/panels/local-model/DeviceCapabilitySection.tsx` around lines 45 - 51, Remove the dead install-state contract by deleting the unused install-related constants and any conditional branches that reference them: remove void onTriggerOllamaInstall, void isTriggeringInstall, void installState, void installWarning, void installError and the const installInProgress/installFailed flags, then eliminate any UI paths or prop wiring that check those symbols (e.g., install retry buttons, install warnings, and install error displays) and the prop types/signatures that still accept install props so the component and its parent no longer expect or pass these obsolete install values (refer to DeviceCapabilitySection and any handlers named onTriggerOllamaInstall/isTriggeringInstall).app/test/e2e/specs/local-model-runtime.spec.ts (1)
46-49: ⚡ Quick winAdd an explicit re-enable tracker for this skipped suite.
describe.skiphere can silently become permanent. Please add a linked TODO/issue with concrete unblock criteria for re-enabling this flow once the harness is available.🤖 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 `@app/test/e2e/specs/local-model-runtime.spec.ts` around lines 46 - 49, Add an explicit re-enable tracker comment immediately above the describe.skip('Local model runtime flow') declaration: create or link to a TODO/GitHub issue ID and include clear unblock criteria (e.g., "deterministic mockable local-runtime harness available for WDIO"), an owner, and a target milestone or PR that will re-enable the test; ensure the comment references describe.skip so reviewers can find it easily and update the issue link/ID when the harness exists.app/src/components/settings/panels/local-model/ModelStatusSection.tsx (1)
74-87: ⚡ Quick winRemove the now-obsolete install/path/repair props from this component API.
The props are kept only to be voided, which preserves stale call-site requirements and obscures the new external-runtime behavior. Tightening the prop surface here will simplify callers and reduce drift.
🤖 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 `@app/src/components/settings/panels/local-model/ModelStatusSection.tsx` around lines 74 - 87, Remove the obsolete install/path/repair props from the ModelStatusSection component API by deleting the unused prop declarations and usages for isTriggeringDownload, bootstrapMessage, isInstalling, isInstallError, showErrorDetail, ollamaPathInput, isSettingPath, runtimeEnabled, onTriggerDownload, onSetOllamaPath, onClearOllamaPath, onSetOllamaPathInput, onToggleErrorDetail, and onRepairAction; update the component's props type/interface and its call sites to stop passing these props (use the new external-runtime behavior instead), and remove any internal references or conditional logic that only existed to support them so the component surface is tightened and callers no longer need to supply voided props.app/src/components/settings/panels/local-model/ModelStatusSection.test.tsx (1)
172-241: ⚡ Quick winExtract a shared
downloadsfixture for the unavailable-Ollama tests.The same large
downloadsobject is duplicated across two tests, which makes schema updates error-prone.♻️ Suggested cleanup
+const makeUnavailableOllamaDownloads = () => ({ + state: 'idle', + warning: null, + progress: 0, + downloaded_bytes: null, + total_bytes: null, + speed_bps: null, + eta_seconds: null, + ollama_available: false, + chat: { id: 'gemma3:1b-it-qat', provider: 'ollama', state: 'missing', progress: null, downloaded_bytes: null, total_bytes: null, speed_bps: null, eta_seconds: null, warning: null, path: null }, + vision: { id: '', provider: 'ollama', state: 'missing', progress: null, downloaded_bytes: null, total_bytes: null, speed_bps: null, eta_seconds: null, warning: null, path: null }, + embedding: { id: 'bge-m3', provider: 'ollama', state: 'missing', progress: null, downloaded_bytes: null, total_bytes: null, speed_bps: null, eta_seconds: null, warning: null, path: null }, + stt: { id: 'whisper', provider: 'whisper', state: 'missing', progress: null, downloaded_bytes: null, total_bytes: null, speed_bps: null, eta_seconds: null, warning: null, path: null }, + tts: { id: 'piper', provider: 'piper', state: 'missing', progress: null, downloaded_bytes: null, total_bytes: null, speed_bps: null, eta_seconds: null, warning: null, path: null }, +}); ... - downloads={{ ...large object... }} + downloads={makeUnavailableOllamaDownloads()} ... - downloads={{ ...large object... }} + downloads={makeUnavailableOllamaDownloads()}Also applies to: 254-323
🤖 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 `@app/src/components/settings/panels/local-model/ModelStatusSection.test.tsx` around lines 172 - 241, Duplicate large downloads object used in the unavailable-Ollama tests should be extracted to a shared fixture to avoid drift; create a single named constant (e.g., unavailableOllamaDownloads) exported/defined near the top of ModelStatusSection.test.tsx and replace the two inline downloads objects in the tests with that constant (references: the downloads object literal in the tests and the test cases that assert unavailable Ollama behavior). Ensure the fixture preserves all keys shown (state, chat, vision, embedding, stt, tts, etc.) and update both places (the duplicate at lines ~172-241 and the similar block at ~254-323) to reference the new constant.
🤖 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.
Inline comments:
In `@src/openhuman/inference/ops_tests.rs`:
- Around line 4-12: The helper disabled_config() currently creates a TempDir
that gets dropped on return causing workspace_dir and config_path to point to a
deleted location; change disabled_config() to return both the Config and the
TempDir (e.g., (Config, TempDir)) so the TempDir lives for the duration of each
test, update all tests that call disabled_config() to use the (config, _tmp)
pattern, and ensure you set config.workspace_dir and config.config_path from
tmp.path() before returning the tuple (referencing disabled_config,
config.workspace_dir, and config.config_path).
In `@src/openhuman/inference/ops.rs`:
- Around line 11-85: Add debug/trace logging to each inference delegate wrapper
(inference_status, inference_summarize, inference_prompt,
inference_vision_prompt, inference_embed, inference_chat,
inference_should_react, inference_analyze_sentiment, inference_should_send_gif,
inference_tenor_search): log entry with key inputs, log exit with outcome
(success path) and any important return metadata, and log errors/failures
including the error string from the awaited call (local_ai::rpc::... or
local_ai::sentiment::..., local_ai::gif_decision::...). Use tracing::debug or
tracing::trace at entry/exit and tracing::error when the Result is Err, and
avoid logging sensitive data (e.g., redact full messages if needed).
In `@src/openhuman/local_ai/ops.rs`:
- Around line 351-354: Replace the capability-specific success message "local ai
voice asset download triggered" used in the RpcOutcome::single_log call with a
capability-neutral message (e.g., "local ai asset download triggered" or include
a dynamic type if an asset_type/variant is available) so it applies to
chat/vision/embedding assets as well; locate the RpcOutcome::single_log(output,
"...") invocation in the function that returns Ok(RpcOutcome::single_log(...))
and update the string literal to a neutral wording or interpolate the asset type
variable if present.
In `@src/openhuman/local_ai/service/bootstrap.rs`:
- Around line 304-309: In the Err arm handling
ensure_ollama_server(&effective_config).await, add an explicit error log before
mutating status and returning: log a development-oriented message (e.g.,
"bootstrap failure: cannot connect to runtime") that includes the error (err)
and relevant context from effective_config so traces show why
ensure_ollama_server failed; place this log call immediately before obtaining
self.status.lock(), then leave the existing status.state, status.error_category,
and status.warning assignment (which uses format_degraded_warning) and the
return unchanged.
---
Outside diff comments:
In `@src/openhuman/local_ai/README.md`:
- Line 3: Update the module synopsis in the README for the local_ai module to
remove claims that it "owns the bundled Ollama runtime" and "download + install
management" and instead state that external-runtime now owns the bundled
runtimes and asset install/management; keep references to what local_ai still
owns (e.g., LM Studio integration, whisper.cpp, Piper, sentiment scoring,
vision-embedding routing, GIF heuristic, and the per-session LocalAiService
singleton) and clarify that remote-provider HTTP transport (providers/) and
agent tool loop (agent/) remain out of scope.
In `@src/openhuman/local_ai/service/ollama_admin.rs`:
- Around line 46-52: The doc comment for ensure_ollama_server_fresh is stale:
the function currently just delegates to ensure_ollama_server but the comment
claims it forces a fresh install; either update the comment to reflect
delegation or implement the fresh-install behavior. To fix, either (A) change
the doc for ensure_ollama_server_fresh to state that it currently calls
ensure_ollama_server and does not force reinstall, or (B) implement the intended
behavior by making ensure_ollama_server_fresh clear any cached/workspace
binaries (or call an existing install helper with a "force" flag) and then call
the install/start routines before returning; locate ensure_ollama_server_fresh
and ensure_ollama_server to add the cache-clear or force-install logic or to
update the comment accordingly.
---
Nitpick comments:
In `@app/src/components/settings/panels/local-model/DeviceCapabilitySection.tsx`:
- Around line 45-51: Remove the dead install-state contract by deleting the
unused install-related constants and any conditional branches that reference
them: remove void onTriggerOllamaInstall, void isTriggeringInstall, void
installState, void installWarning, void installError and the const
installInProgress/installFailed flags, then eliminate any UI paths or prop
wiring that check those symbols (e.g., install retry buttons, install warnings,
and install error displays) and the prop types/signatures that still accept
install props so the component and its parent no longer expect or pass these
obsolete install values (refer to DeviceCapabilitySection and any handlers named
onTriggerOllamaInstall/isTriggeringInstall).
In `@app/src/components/settings/panels/local-model/ModelStatusSection.test.tsx`:
- Around line 172-241: Duplicate large downloads object used in the
unavailable-Ollama tests should be extracted to a shared fixture to avoid drift;
create a single named constant (e.g., unavailableOllamaDownloads)
exported/defined near the top of ModelStatusSection.test.tsx and replace the two
inline downloads objects in the tests with that constant (references: the
downloads object literal in the tests and the test cases that assert unavailable
Ollama behavior). Ensure the fixture preserves all keys shown (state, chat,
vision, embedding, stt, tts, etc.) and update both places (the duplicate at
lines ~172-241 and the similar block at ~254-323) to reference the new constant.
In `@app/src/components/settings/panels/local-model/ModelStatusSection.tsx`:
- Around line 74-87: Remove the obsolete install/path/repair props from the
ModelStatusSection component API by deleting the unused prop declarations and
usages for isTriggeringDownload, bootstrapMessage, isInstalling, isInstallError,
showErrorDetail, ollamaPathInput, isSettingPath, runtimeEnabled,
onTriggerDownload, onSetOllamaPath, onClearOllamaPath, onSetOllamaPathInput,
onToggleErrorDetail, and onRepairAction; update the component's props
type/interface and its call sites to stop passing these props (use the new
external-runtime behavior instead), and remove any internal references or
conditional logic that only existed to support them so the component surface is
tightened and callers no longer need to supply voided props.
In `@app/test/e2e/specs/local-model-runtime.spec.ts`:
- Around line 46-49: Add an explicit re-enable tracker comment immediately above
the describe.skip('Local model runtime flow') declaration: create or link to a
TODO/GitHub issue ID and include clear unblock criteria (e.g., "deterministic
mockable local-runtime harness available for WDIO"), an owner, and a target
milestone or PR that will re-enable the test; ensure the comment references
describe.skip so reviewers can find it easily and update the issue link/ID when
the harness exists.
In `@src/openhuman/local_ai/service/ollama_admin_tests.rs`:
- Around line 126-219: The test module has grown too large; move the three new
tests (ensure_ollama_server_requires_external_runtime_when_unreachable,
ensure_ollama_server_reports_broken_external_runner_without_restart_attempt,
assets_status_marks_ollama_unavailable_when_runtime_is_down_even_if_binary_exists)
into a new sibling test module file to reduce module size and merge-conflict
surface. Create a separate test-only module file, import the same helpers (e.g.,
local_ai_test_guard, spawn_mock, Config, LocalAiService) and any env
setup/teardown logic, copy those three test functions verbatim into it, and
remove them from the original file; ensure the new file compiles by adding the
necessary use statements and keeping the #[tokio::test] annotations intact.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: deaef292-c2ea-4738-8474-77952bd66846
📒 Files selected for processing (39)
app/src/components/settings/panels/LocalModelDebugPanel.tsxapp/src/components/settings/panels/local-model/DeviceCapabilitySection.test.tsxapp/src/components/settings/panels/local-model/DeviceCapabilitySection.tsxapp/src/components/settings/panels/local-model/ModelDownloadSection.test.tsxapp/src/components/settings/panels/local-model/ModelDownloadSection.tsxapp/src/components/settings/panels/local-model/ModelStatusSection.test.tsxapp/src/components/settings/panels/local-model/ModelStatusSection.tsxapp/src/pages/onboarding/steps/LocalAIStep.tsxapp/src/services/api/__tests__/aiSettingsApi.test.tsapp/src/services/api/aiSettingsApi.tsapp/src/utils/__tests__/localAiBootstrap.test.tsapp/src/utils/localAiBootstrap.tsapp/src/utils/tauriCommands/localAi.tsapp/test/e2e/specs/local-model-runtime.spec.tssrc/core/all.rssrc/core/cli_tests.rssrc/core/jsonrpc_tests.rssrc/openhuman/app_state/ops.rssrc/openhuman/channels/providers/presentation.rssrc/openhuman/inference/mod.rssrc/openhuman/inference/ops.rssrc/openhuman/inference/ops_tests.rssrc/openhuman/inference/schemas.rssrc/openhuman/inference/schemas_tests.rssrc/openhuman/local_ai/README.mdsrc/openhuman/local_ai/mod.rssrc/openhuman/local_ai/ops.rssrc/openhuman/local_ai/schemas.rssrc/openhuman/local_ai/schemas_tests.rssrc/openhuman/local_ai/service/assets.rssrc/openhuman/local_ai/service/bootstrap.rssrc/openhuman/local_ai/service/ollama_admin.rssrc/openhuman/local_ai/service/ollama_admin_tests.rssrc/openhuman/local_ai/service/public_infer.rssrc/openhuman/local_ai/service/public_infer_tests.rssrc/openhuman/local_ai/types.rssrc/openhuman/mod.rssrc/openhuman/subconscious/executor.rstests/json_rpc_e2e.rs
💤 Files with no reviewable changes (4)
- app/src/services/api/aiSettingsApi.ts
- src/core/jsonrpc_tests.rs
- src/openhuman/local_ai/schemas_tests.rs
- src/openhuman/local_ai/schemas.rs
…/inference/ Move local_ai/, providers/, and voice inference files under inference/ so that the entire inference surface lives in one place: inference/local/ ← Ollama / LM Studio / Whisper / Piper runtime inference/provider/ ← cloud + local provider trait, routing, reliability inference/voice/ ← STT and TTS inference implementations inference/http/ ← OpenAI-compatible /v1/chat/completions + /v1/models src/openhuman/local_ai/ and src/openhuman/providers/ become thin re-export shims for backward compatibility; src/openhuman/voice/ re-exports the moved modules. Add OpenAI-compatible HTTP endpoint mounted at /v1 with three integration tests (no-bearer → 401, bearer present → not 401/403, GET /v1/models → 401 without bearer). Both cargo check manifests (core + Tauri shell) pass clean.
…cker wiring - temperature_unsupported_models config field (glob, default ["o1*","o3*","o4*","gpt-5*"]) added to Config; serde default and Config::default() both initialize it - temperature.rs: glob_match helper + temperature_for_model; 13 unit tests cover prefix/suffix/contains globs and all default model patterns - ApiChatRequest and NativeChatRequest temperature fields changed to Option<f64> with skip_serializing_if, so matched models get no temperature key on the wire - OpenAiCompatibleProvider gains temperature_unsupported_models field and effective_temperature() helper; all six request-body construction sites updated - Factory threads the config list through make_cloud_provider_by_slug and make_ollama_provider via new with_temperature_unsupported_models builder - server.rs logs a warning when the caller supplies temperature for an unsupported model - tests/inference_provider_e2e.rs: 14 tests via wiremock covering OpenAI-compat chat, temperature present/absent, Anthropic auth style, streaming, Ollama, /v1 HTTP auth - scripts/test-rust-inference-e2e.sh + e2e/docker-compose.yml inference-e2e service
Contributor
There was a problem hiding this comment.
Actionable comments posted: 6
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (2)
src/openhuman/inference/voice/local_transcribe.rs (1)
18-23:⚠️ Potential issue | 🟡 Minor | ⚡ Quick winFix stale module path in voice-install docs.
Line 23 still references
local_ai/paths.rs; after this refactor it should point to the inference paths module.📝 Proposed doc fix
-//! `local_ai/paths.rs` picks it up automatically — no env var to set. +//! `inference/paths.rs` picks it up automatically — no env var to set.🤖 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/inference/voice/local_transcribe.rs` around lines 18 - 23, Update the stale docs reference that currently points to local_ai/paths.rs: change the module path to the new inference paths module so the comment about install_whisper (crate::openhuman::inference::local::install_whisper) points to the correct resolver (resolve_whisper_binary) in the inference paths module (e.g., crate::openhuman::inference::paths::resolve_whisper_binary) so readers can find the helper after the refactor.src/openhuman/inference/voice/local_speech.rs (1)
29-42:⚠️ Potential issue | 🟡 Minor | ⚡ Quick winFix stale module reference in the Piper docs.
Line 42 still points to
local_ai/paths.rs, but this flow now resolves via the inference paths module.📝 Proposed doc fix
-//! helper in `local_ai/paths.rs` picks it up automatically. +//! helper in `src/openhuman/inference/paths.rs` picks it up automatically.🤖 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/inference/voice/local_speech.rs` around lines 29 - 42, Update the stale doc link that currently points to local_ai/paths.rs so it references the inference paths helper instead: replace the mention of local_ai/paths.rs with the fully-qualified crate::openhuman::inference::paths::resolve_piper_binary (and ensure surrounding text still references the installer function crate::openhuman::inference::local::install_piper and the same install flow), so readers are directed to the correct resolve_piper_binary symbol in the inference paths module.
🧹 Nitpick comments (2)
src/openhuman/inference/local/service/ollama_admin.rs (1)
42-45: ⚡ Quick winLog unreachable external-runtime failures before returning.
This changed failure path returns a user-facing error but emits no log line with endpoint context. Add a warning log so production failures are grep-visible.
As per coding guidelines: "Ship heavy `debug`/`trace` level logs on new flows and critical checkpoints ... and error paths."📈 Suggested fix
let base_url = ollama_base_url(); + log::warn!( + "[local_ai] external Ollama runtime unreachable at {}", + base_url + ); Err(format!( "OpenHuman no longer starts or installs Ollama automatically. Start your inference runtime yourself and make sure it is reachable at {base_url}." ))🤖 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/inference/local/service/ollama_admin.rs` around lines 42 - 45, The current error path builds and returns an Err with the ollama_base_url but does not emit any log; before returning the Err in the Ollama admin flow (the spot that calls ollama_base_url() and returns Err(format!(...))), add a warning-level log that includes the same base_url and a brief context string so production failures are grep-visible (use the project's logging macro, e.g. warn! or tracing::warn!, to log the endpoint and message) and then return the Err as before.src/openhuman/inference/schemas.rs (1)
213-389: 💤 Low valueConsider returning
Resultinstead of panicking for unknown schema names.The
panic!at line 387 will crash the process ifschemas()is ever called with an unrecognized function name. While current internal usage viaall_controller_schemas()andall_registered_controllers()uses hardcoded valid strings, this function ispuband could be called by other code with dynamic input.♻️ Suggested defensive alternative
-pub fn schemas(function: &str) -> ControllerSchema { +pub fn schemas(function: &str) -> Option<ControllerSchema> { match function { - "status" => ControllerSchema { + "status" => Some(ControllerSchema { namespace: "inference", function: "status", // ... rest unchanged, wrap each arm in Some(...) - }, + }), // ... other arms - other => panic!("unknown inference schema: {other}"), + _ => None, } }Then update callers to
.expect("known schema")or handleNone.🤖 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/inference/schemas.rs` around lines 213 - 389, The schemas() function currently panics on unknown names (panic!("unknown inference schema: {other}")), which can crash the process; change its signature to return Result<ControllerSchema, SchemaError> or Option<ControllerSchema> (e.g., pub fn schemas(function: &str) -> Option<ControllerSchema>) and replace the final panic arm with None or Err(...), then update callers such as all_controller_schemas() and all_registered_controllers() to handle the None/Err (e.g., .expect("known schema") where appropriate or propagate the error) so unknown schema lookups are handled defensively instead of aborting.
🤖 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.
Inline comments:
In `@app/src/services/api/aiSettingsApi.ts`:
- Around line 272-275: Replace the direct callCoreRpc call in aiSettingsApi (the
call that invokes 'openhuman.inference_list_models' with params { provider_id:
providerId }) with the Tauri IPC wrapper used elsewhere: call
tauriCommands.invoke('core_rpc_relay', { method:
'openhuman.inference_list_models', params: { provider_id: providerId } }) and
unwrap the returned payload into the same shape ({ result: { models: ModelInfo[]
} }) so the rest of the function continues to operate unchanged; preserve
existing try/catch logic but rely on the invoke path so the call waits for the
in-process core relay instead of using callCoreRpc.
In `@src/openhuman/inference/local/service/ollama_admin.rs`:
- Around line 48-55: The docstring for ensure_ollama_server_fresh is stale: it
claims to force a fresh install/start but the function simply calls
ensure_ollama_server; either restore the fresh-install behavior or update the
docs. Fix by editing the function ensure_ollama_server_fresh: if you intend to
keep delegation, change the docstring to state that it currently delegates to
ensure_ollama_server and acts as an alias/retry wrapper; if you intend to
reinstate fresh behavior, implement the fresh-install logic (e.g., bypass
cache/workspace and force reinstall/start) within ensure_ollama_server_fresh
before calling or instead of calling ensure_ollama_server. Reference
ensure_ollama_server_fresh and ensure_ollama_server when making the change so
callers and reviewers see the intended contract.
In `@src/openhuman/inference/local/service/public_infer.rs`:
- Around line 11-16: The error message in external_ollama_request_error
currently embeds ollama_base_url() verbatim and may leak credentials; instead
parse the URL returned by ollama_base_url(), remove or replace any userinfo
(username/password) before formatting the message, and include the sanitized URL
string; locate external_ollama_request_error and use a URL-parsing utility
(e.g., url::Url::parse) to strip userinfo or substitute a placeholder like
"<redacted>" prior to including the value in the formatted error string.
In `@src/openhuman/inference/provider/ops.rs`:
- Around line 86-87: The match arm in ops.rs currently treats
AuthStyle::OpenhumanJwt like AuthStyle::None, sending unauthenticated requests;
update the match so AuthStyle::OpenhumanJwt follows the authenticated path used
for other token-based styles (i.e., call the same code that attaches the
JWT/bearer header or signs the request instead of returning plain request).
Locate the match on AuthStyle (the arm returning `request`) and change the
OpenhumanJwt branch to invoke the existing request-authentication helper (the
method that adds Authorization headers or signs requests) so model-list calls
include the provider JWT.
- Around line 96-102: The code currently returns a raw/truncated upstream body;
instead, call the module's secret-scrubbing helper before truncation and
returning the error. Replace use of body/truncated with a sanitized string
produced by e.g. crate::openhuman::util::scrub_secrets(&body) (or the module's
actual secret-scrubbing function), then pass that sanitized text into
crate::openhuman::util::truncate_with_ellipsis and include the
sanitized+truncated result in the Err message (keeping status.as_u16() as-is).
In `@tests/json_rpc_e2e.rs`:
- Around line 3626-3808: Add JSON-RPC e2e assertions for the new methods: call
openhuman.inference_list_models (via post_json_rpc) after configuring the
provider (e.g., in json_rpc_inference_namespace_lm_studio_prompt_and_status) and
assert_no_jsonrpc_error and that the returned list contains the expected
"local-model" entry; also call openhuman.inference_update_model_settings (via
post_json_rpc) with a payload that changes model settings and
assert_no_jsonrpc_error and that a subsequent inference_status or
inference_list_models reflects the updated setting. Locate the test flow around
post_json_rpc calls and helpers like assert_no_jsonrpc_error,
assert_jsonrpc_error, extract_string_outcome, and add these new calls and
assertions in the same tests so the RPC transport and parameter shapes for
inference_list_models and inference_update_model_settings are validated.
---
Outside diff comments:
In `@src/openhuman/inference/voice/local_speech.rs`:
- Around line 29-42: Update the stale doc link that currently points to
local_ai/paths.rs so it references the inference paths helper instead: replace
the mention of local_ai/paths.rs with the fully-qualified
crate::openhuman::inference::paths::resolve_piper_binary (and ensure surrounding
text still references the installer function
crate::openhuman::inference::local::install_piper and the same install flow), so
readers are directed to the correct resolve_piper_binary symbol in the inference
paths module.
In `@src/openhuman/inference/voice/local_transcribe.rs`:
- Around line 18-23: Update the stale docs reference that currently points to
local_ai/paths.rs: change the module path to the new inference paths module so
the comment about install_whisper
(crate::openhuman::inference::local::install_whisper) points to the correct
resolver (resolve_whisper_binary) in the inference paths module (e.g.,
crate::openhuman::inference::paths::resolve_whisper_binary) so readers can find
the helper after the refactor.
---
Nitpick comments:
In `@src/openhuman/inference/local/service/ollama_admin.rs`:
- Around line 42-45: The current error path builds and returns an Err with the
ollama_base_url but does not emit any log; before returning the Err in the
Ollama admin flow (the spot that calls ollama_base_url() and returns
Err(format!(...))), add a warning-level log that includes the same base_url and
a brief context string so production failures are grep-visible (use the
project's logging macro, e.g. warn! or tracing::warn!, to log the endpoint and
message) and then return the Err as before.
In `@src/openhuman/inference/schemas.rs`:
- Around line 213-389: The schemas() function currently panics on unknown names
(panic!("unknown inference schema: {other}")), which can crash the process;
change its signature to return Result<ControllerSchema, SchemaError> or
Option<ControllerSchema> (e.g., pub fn schemas(function: &str) ->
Option<ControllerSchema>) and replace the final panic arm with None or Err(...),
then update callers such as all_controller_schemas() and
all_registered_controllers() to handle the None/Err (e.g., .expect("known
schema") where appropriate or propagate the error) so unknown schema lookups are
handled defensively instead of aborting.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: 29c390c6-cb6a-425c-b040-cba30bb58fd6
⛔ Files ignored due to path filters (1)
Cargo.lockis excluded by!**/*.lock
📒 Files selected for processing (197)
Cargo.tomlapp/src/components/settings/panels/local-model/ModelDownloadSection.tsxapp/src/services/__tests__/coreRpcClient.test.tsapp/src/services/__tests__/rpcMethods.test.tsapp/src/services/api/__tests__/aiSettingsApi.test.tsapp/src/services/api/aiSettingsApi.tsapp/src/services/rpcMethods.tsapp/src/utils/__tests__/tauriCommands.test.tsapp/src/utils/tauriCommands/__tests__/config.test.tsapp/src/utils/tauriCommands/config.test.tsapp/src/utils/tauriCommands/config.tsapp/src/utils/tauriCommands/localAi.tse2e/docker-compose.ymlgitbooks/developing/e2e-testing.mdscripts/test-rust-inference-e2e.shsrc/api/rest.rssrc/core/all.rssrc/core/jsonrpc.rssrc/core/legacy_aliases.rssrc/core/observability.rssrc/main.rssrc/openhuman/agent/bus.rssrc/openhuman/agent/cost.rssrc/openhuman/agent/dispatcher.rssrc/openhuman/agent/dispatcher_tests.rssrc/openhuman/agent/harness/bughunt_tests.rssrc/openhuman/agent/harness/fork_context.rssrc/openhuman/agent/harness/harness_gap_tests.rssrc/openhuman/agent/harness/parse.rssrc/openhuman/agent/harness/session/builder.rssrc/openhuman/agent/harness/session/runtime.rssrc/openhuman/agent/harness/session/runtime_tests.rssrc/openhuman/agent/harness/session/tests.rssrc/openhuman/agent/harness/session/transcript.rssrc/openhuman/agent/harness/session/turn.rssrc/openhuman/agent/harness/session/turn_tests.rssrc/openhuman/agent/harness/session/types.rssrc/openhuman/agent/harness/subagent_runner/extract_tool.rssrc/openhuman/agent/harness/subagent_runner/mod.rssrc/openhuman/agent/harness/subagent_runner/ops.rssrc/openhuman/agent/harness/subagent_runner/ops_tests.rssrc/openhuman/agent/harness/test_support.rssrc/openhuman/agent/harness/test_support_test.rssrc/openhuman/agent/harness/tests.rssrc/openhuman/agent/harness/tool_loop.rssrc/openhuman/agent/harness/tool_loop_tests.rssrc/openhuman/agent/multimodal.rssrc/openhuman/agent/schemas.rssrc/openhuman/agent/stop_hooks.rssrc/openhuman/agent/tests.rssrc/openhuman/agent/triage/evaluator.rssrc/openhuman/agent/triage/evaluator_tests.rssrc/openhuman/agent/triage/routing.rssrc/openhuman/app_state/ops.rssrc/openhuman/autocomplete/core/engine.rssrc/openhuman/channels/context.rssrc/openhuman/channels/providers/web.rssrc/openhuman/channels/routes.rssrc/openhuman/channels/routes_tests.rssrc/openhuman/channels/runtime/dispatch.rssrc/openhuman/channels/runtime/startup.rssrc/openhuman/channels/tests/common.rssrc/openhuman/channels/tests/context.rssrc/openhuman/channels/tests/discord_integration.rssrc/openhuman/channels/tests/memory.rssrc/openhuman/channels/tests/runtime_dispatch.rssrc/openhuman/channels/tests/runtime_tool_calls.rssrc/openhuman/channels/tests/telegram_integration.rssrc/openhuman/config/ops.rssrc/openhuman/config/schema/cloud_providers.rssrc/openhuman/config/schema/load.rssrc/openhuman/config/schema/types.rssrc/openhuman/config/schemas.rssrc/openhuman/context/guard.rssrc/openhuman/context/manager.rssrc/openhuman/context/manager_tests.rssrc/openhuman/context/microcompact.rssrc/openhuman/context/pipeline.rssrc/openhuman/context/summarizer.rssrc/openhuman/context/summarizer_tests.rssrc/openhuman/credentials/ops.rssrc/openhuman/cron/scheduler.rssrc/openhuman/doctor/core.rssrc/openhuman/embeddings/cloud.rssrc/openhuman/embeddings/factory.rssrc/openhuman/inference/device.rssrc/openhuman/inference/http/mod.rssrc/openhuman/inference/http/server.rssrc/openhuman/inference/http/tests.rssrc/openhuman/inference/http/types.rssrc/openhuman/inference/local/core.rssrc/openhuman/inference/local/install.rssrc/openhuman/inference/local/install_piper.rssrc/openhuman/inference/local/install_whisper.rssrc/openhuman/inference/local/lm_studio.rssrc/openhuman/inference/local/mod.rssrc/openhuman/inference/local/ollama.rssrc/openhuman/inference/local/ops.rssrc/openhuman/inference/local/ops_tests.rssrc/openhuman/inference/local/process_util.rssrc/openhuman/inference/local/provider.rssrc/openhuman/inference/local/schemas.rssrc/openhuman/inference/local/schemas_tests.rssrc/openhuman/inference/local/service/assets.rssrc/openhuman/inference/local/service/bootstrap.rssrc/openhuman/inference/local/service/lm_studio.rssrc/openhuman/inference/local/service/mod.rssrc/openhuman/inference/local/service/ollama_admin.rssrc/openhuman/inference/local/service/ollama_admin_tests.rssrc/openhuman/inference/local/service/public_infer.rssrc/openhuman/inference/local/service/public_infer_tests.rssrc/openhuman/inference/local/service/spawn_marker.rssrc/openhuman/inference/local/service/speech.rssrc/openhuman/inference/local/service/vision_embed.rssrc/openhuman/inference/local/service/whisper_engine.rssrc/openhuman/inference/local/voice_install_common.rssrc/openhuman/inference/mod.rssrc/openhuman/inference/model_ids.rssrc/openhuman/inference/ops.rssrc/openhuman/inference/ops_tests.rssrc/openhuman/inference/parse.rssrc/openhuman/inference/paths.rssrc/openhuman/inference/presets.rssrc/openhuman/inference/presets_tests.rssrc/openhuman/inference/provider/billing_error.rssrc/openhuman/inference/provider/compatible.rssrc/openhuman/inference/provider/compatible_dump.rssrc/openhuman/inference/provider/compatible_parse.rssrc/openhuman/inference/provider/compatible_stream.rssrc/openhuman/inference/provider/compatible_tests.rssrc/openhuman/inference/provider/compatible_types.rssrc/openhuman/inference/provider/factory.rssrc/openhuman/inference/provider/factory_test.rssrc/openhuman/inference/provider/mod.rssrc/openhuman/inference/provider/openhuman_backend.rssrc/openhuman/inference/provider/ops.rssrc/openhuman/inference/provider/reliable.rssrc/openhuman/inference/provider/reliable_tests.rssrc/openhuman/inference/provider/router.rssrc/openhuman/inference/provider/router_test.rssrc/openhuman/inference/provider/schemas.rssrc/openhuman/inference/provider/temperature.rssrc/openhuman/inference/provider/thread_context.rssrc/openhuman/inference/provider/traits.rssrc/openhuman/inference/provider/traits_tests.rssrc/openhuman/inference/schemas.rssrc/openhuman/inference/schemas_tests.rssrc/openhuman/inference/sentiment.rssrc/openhuman/inference/types.rssrc/openhuman/inference/voice/cloud_transcribe.rssrc/openhuman/inference/voice/hallucination.rssrc/openhuman/inference/voice/local_speech.rssrc/openhuman/inference/voice/local_transcribe.rssrc/openhuman/inference/voice/mod.rssrc/openhuman/inference/voice/postprocess.rssrc/openhuman/inference/voice/streaming.rssrc/openhuman/learning/linkedin_enrichment.rssrc/openhuman/learning/reflection.rssrc/openhuman/learning/reflection_tests.rssrc/openhuman/learning/transcript_ingest/extract.rssrc/openhuman/learning/transcript_ingest/tests.rssrc/openhuman/local_ai/README.mdsrc/openhuman/local_ai/mod.rssrc/openhuman/local_ai/schemas.rssrc/openhuman/mcp_server/tools.rssrc/openhuman/memory/store/factories.rssrc/openhuman/memory/tree/chat/cloud.rssrc/openhuman/memory/tree/chat/mod.rssrc/openhuman/memory/tree/score/embed/factory.rssrc/openhuman/migrations/mod_tests.rssrc/openhuman/migrations/phase_out_profile_md_tests.rssrc/openhuman/migrations/unify_ai_provider_settings.rssrc/openhuman/mod.rssrc/openhuman/providers/schemas.rssrc/openhuman/routing/factory.rssrc/openhuman/routing/mod.rssrc/openhuman/routing/provider.rssrc/openhuman/routing/provider_tests.rssrc/openhuman/screen_intelligence/processing_worker.rssrc/openhuman/subconscious/executor.rssrc/openhuman/threads/ops.rssrc/openhuman/tools/impl/agent/delegate.rssrc/openhuman/tools/impl/agent/spawn_parallel_agents_test.rssrc/openhuman/tools/impl/agent/spawn_worker_thread.rssrc/openhuman/tools/impl/agent/todo_write.rssrc/openhuman/tools/ops.rssrc/openhuman/tree_summarizer/engine.rssrc/openhuman/tree_summarizer/ops.rssrc/openhuman/voice/mod.rssrc/openhuman/voice/ops.rssrc/openhuman/voice/types.rstests/agent_builder_public.rstests/agent_harness_public.rstests/agent_multimodal_public.rstests/calendar_grounding_e2e.rstests/inference_provider_e2e.rstests/json_rpc_e2e.rs
💤 Files with no reviewable changes (4)
- src/openhuman/providers/schemas.rs
- src/openhuman/local_ai/mod.rs
- src/openhuman/local_ai/README.md
- src/openhuman/local_ai/schemas.rs
✅ Files skipped from review due to trivial changes (32)
- src/openhuman/inference/http/mod.rs
- scripts/test-rust-inference-e2e.sh
- src/openhuman/embeddings/cloud.rs
- src/openhuman/agent/harness/subagent_runner/mod.rs
- app/src/utils/tests/tauriCommands.test.ts
- src/openhuman/config/schema/cloud_providers.rs
- src/openhuman/agent/harness/fork_context.rs
- src/openhuman/autocomplete/core/engine.rs
- src/openhuman/context/manager.rs
- src/openhuman/inference/provider/thread_context.rs
- src/openhuman/agent/multimodal.rs
- gitbooks/developing/e2e-testing.md
- src/openhuman/inference/local/service/mod.rs
- src/openhuman/context/manager_tests.rs
- src/openhuman/context/pipeline.rs
- src/openhuman/learning/transcript_ingest/tests.rs
- src/openhuman/agent/harness/test_support_test.rs
- src/openhuman/context/microcompact.rs
- src/openhuman/migrations/phase_out_profile_md_tests.rs
- src/main.rs
- src/openhuman/agent/harness/session/transcript.rs
- src/openhuman/agent/harness/harness_gap_tests.rs
- src/openhuman/routing/mod.rs
- src/openhuman/migrations/unify_ai_provider_settings.rs
- src/openhuman/agent/harness/parse.rs
- src/openhuman/migrations/mod_tests.rs
- tests/agent_harness_public.rs
- src/openhuman/channels/tests/common.rs
- src/openhuman/memory/tree/chat/cloud.rs
- src/openhuman/tools/impl/agent/todo_write.rs
- src/openhuman/inference/local/mod.rs
- src/openhuman/tools/impl/agent/spawn_parallel_agents_test.rs
🚧 Files skipped from review as they are similar to previous changes (4)
- src/openhuman/mod.rs
- app/src/services/api/tests/aiSettingsApi.test.ts
- src/openhuman/app_state/ops.rs
- src/openhuman/inference/schemas_tests.rs
- ollama_admin: correct stale docstring on `ensure_ollama_server_fresh` to reflect external-runtime mode (no auto install/start anymore) - public_infer: redact userinfo/query/fragment from ollama_base_url before embedding in error payloads; add unit tests - inference::provider::ops: route AuthStyle::OpenhumanJwt through the Bearer header path instead of falling into the unauthenticated branch - inference::provider::ops: sanitize upstream provider error bodies via sanitize_api_error before returning them in RPC errors - tests/json_rpc_e2e: add coverage for openhuman.inference_list_models and openhuman.inference_update_model_settings over the RPC transport
…ernal-ollama-routing
…ernal-ollama-routing # Conflicts: # src/openhuman/agent/harness/session/builder.rs # src/openhuman/tools/impl/agent/todo_write.rs
…guard - inference_diagnostics: return value with empty logs (`RpcOutcome::new`) instead of `single_log` so callers see the diagnostics object directly, matching the legacy `local_ai_diagnostics` shape that the UI and json_rpc_e2e tests assert against (`provider`, `lm_studio_running`, `expected.chat_found`, etc.). - rpcMethods drift guard: read schemas from `inference/provider/schemas.rs` rather than the deleted `providers/schemas.rs`.
…t config `LocalAiService` is a process-wide `OnceCell` singleton whose cached `provider` field is set at first init and never refreshed. After an `inference_update_local_settings` call swaps providers (ollama → lm_studio) the cached value goes stale, so `inference_status` returns the previous provider even though the on-disk config has the new one. CI ran this test as part of the full suite where an earlier ollama-config test had already initialized the singleton, so the lm_studio assertion failed; locally in isolation the singleton picked lm_studio first and the test passed. Overlay the current config's provider on the returned snapshot so `local_ai_status` reflects on-disk config without disturbing other service state (state machine, model ids, etc.).
7 tasks
7 tasks
This was referenced May 18, 2026
AusAgentSmith
pushed a commit
to AusAgentSmith/openhuman
that referenced
this pull request
May 23, 2026
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
src/openhuman/inferencecore module and register a dedicatedopenhuman.inference_*controller surfaceProblem
local_aicatch-all module, which still assumed OpenHuman could install, launch, and manage Ollama directlySolution
inferencedomain in core and rewired callers onto it for status, prompt, summarize, chat, and lightweight inference helpersdownload,download_all_assets,set_ollama_path,shutdown_owned) and simplified onboarding/bootstrap helpers to preset application onlySubmission Checklist
## Related— N/A: no explicit matrix feature IDs were mapped during this refactor passdocs/RELEASE-MANUAL-SMOKE.md) — N/A: no release smoke doc changes required for this code-only refactorCloses #NNNin the## Relatedsection — N/A: no issue number was provided in-sessionImpact
Related
local_ai_*read-path callers to theinference_*namespace over timeAI Authored PR Metadata (required for Codex/Linear PRs)
Linear Issue
Commit & Branch
codex/inference-external-ollama-routing12bed1a522dec88b292f509796eee6e76d75add6Validation Run
pnpm --filter openhuman-app format:checkpnpm typecheckpnpm debug unit src/components/settings/panels/local-model/ModelStatusSection.test.tsx src/utils/localAiBootstrap.test.ts src/pages/onboarding/steps/__tests__/LocalAIStep.test.tsxpnpm debug unit src/services/api/__tests__/aiSettingsApi.test.tspnpm debug unit src/components/settings/panels/local-model/ModelDownloadSection.test.tsxpnpm debug unit src/components/settings/panels/local-model/DeviceCapabilitySection.test.tsxcargo test --manifest-path Cargo.toml inference:: --libcargo test --manifest-path Cargo.toml inference_connection_failure_mentions_external_ollama_runtime --libcargo test --manifest-path Cargo.toml json_rpc_inference_namespace_lm_studio_prompt_and_status --test json_rpc_e2ecargo test --manifest-path Cargo.toml json_rpc_inference_prompt_requires_external_ollama_runtime_when_unreachable --test json_rpc_e2ecargo fmt --manifest-path Cargo.toml --all --check,cargo check --manifest-path Cargo.tomlcargo fmt --manifest-path app/src-tauri/Cargo.toml --all --check,cargo check --manifest-path app/src-tauri/Cargo.tomlValidation Blocked
command:pnpm test:coverageerror:not completed during shipping pass; no final coverage report captured locallyimpact:changed-line coverage confirmation is deferred to CI unless rerun locally in a follow-upBehavior Changes
Parity Contract
inference_*surface plus external-Ollama unreachable and LM Studio happy pathsDuplicate / Superseded PR Handling
Summary by CodeRabbit
New Features
Bug Fixes
Behavior Changes
Tests & Docs