Repository navigation
Refactor canvas utility tools for production reliability - #12
Merged
Merged
Conversation
dovvnloading
added a commit
that referenced
this pull request
Aug 8, 2026
…300) * ADR-016 stage 16.1: structured JSON logging, run_id, log-level setting Structured logging: backend/observability.py's JsonLogFormatter renders one JSON object per line on the rotating file handler (stderr stays plain text for terminal runs); RunRegistry.claim/cancel/_pop (the one choke point every one of the app's 12 dispatch surfaces already shares) now logs "run claimed"/"run cancelled"/"run released" with run_id/kind/ node_id, so every run's lifecycle is traced without touching each surface individually. Print-ban: converted the 5 real print() calls in shipped code (graphlink_settings_store.py x3, graphlink_chat_agent.py, graphlink_token_estimator.py) to logger calls. Ruff's T201 CI enforcement is explicitly deferred to ADR-015/roadmap PR #12 ("ruff + mypy ratchets") - no ruff config or CI step exists yet in this repo, and introducing that pipeline is that PR's own scope, not this ADR's; the underlying state (no print in shipped paths) is met now. Log-level setting: SettingsManager.get_log_level/set_log_level (closed vocabulary DEBUG/INFO/WARNING/ERROR, INFO default), a new setLogLevel app-settings intent that persists AND applies live via apply_log_level(), and a boot-time read in graphlink_desktop.py's main() before configure_logging() attaches handlers. Deliberately NOT wired into backend/app.py's create_app() - that's constructed dozens of times per pytest run and mutating the root logger's level there would silently change what caplog captures for unrelated tests. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> * ADR-016 stage 16.2: user-editable pricing + per-node/session cost UI Backend: SettingsManager.get_pricing_overrides/set_pricing_overrides - an exact-model-id-keyed local pricing table, checked before the built-in prefix table in token_counter.py's estimate_cost_usd. Wired into TokenCounterState via a live accessor (pricing_overrides_fn), so an override change applies to the next reply without a restart. ChatState gains estimated_cost_usd - a point-in-time cost snapshot taken when a reply's usage is stamped (backend/api/intents_chat.py's _on_usage), surviving save/load like prompt_tokens/completion_tokens. TokenCounterState gains session-cumulative totals (sessionPromptTokens/ sessionCompletionTokens/sessionEstimatedCostUsd), accumulated on every real-usage reply. Frontend: ChatNodeView renders a compact "prompt→completion · $cost" badge next to the model badge - data was on the wire since ADR-006 stage 6.8 but never rendered. No settings UI for editing pricing overrides in this sweep, matching the get_mcp_servers/set_mcp_servers precedent from ADR-007 stage 7.5 (data layer now, UI panel deferred). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> * ADR-016 stage 16.3: in-app diagnostics (runs, publish size/rate, provider errors) backend/diagnostics.py adds a per-session DiagnosticsState (recent run outcomes/durations, publish count/bytes/rate, session count) fed by explicit on_claim/on_end callbacks on RunRegistry and a publish-size recorder on SessionBus, plus a process-global bounded deque of provider errors recorded from api_provider.py's single chat-exception translation choke point. Wired into app.py as a new "diagnostics" WS topic. Frontend: DiagnosticsDialog.tsx subscribes to the topic and renders the stats, a recent-runs table, and a provider-errors list, reachable from the app bar next to About/Help. Also fixes a pre-existing CSS overflow bug in the token-counter popout (unbounded width from ADR-006 stage 6.8 growing it to 7 columns without revisiting the layout). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> * ADR-016 stages 16.4 + 16.5: diagnostic bundle, open-log-folder, evals harness Stage 16.4 - backend/diagnostic_bundle.py builds a redacted snapshot (app version, OS, node counts by kind, the diagnostics payload) for issue reports. Redaction is structural: build_diagnostic_bundle's signature takes only (document, diagnostics_state), so no settings manager, log handle, or provider object is reachable from inside it. Exposed as two read-only intents on the existing "diagnostics" topic, with an action row in the Diagnostics dialog that previews and copies the bundle. One field was not content-free by construction: providerErrors[].message started as raw str(exc), which embeds the full absolute path (including the OS username) for an OSError and can carry SDK-echoed request content. record_provider_error now classifies it into a fixed vocabulary of category labels and stores no substring of the original; the full text still reaches graphlink.log. Stage 16.5 - backend/evals/ runs golden fixtures against the chart and structured-output paths, deterministic by default (provider seam patched) and --live against a real provider. The agent-build dimension reports not_implemented rather than a fake pass: ADR-008's Builder loop does not exist yet, so that half of the stage's exit criterion is not met and the ADR now says so. Test plan: full pytest from repo root (2045 passed, 14 skipped); web_ui typecheck, lint, test (1382 passed), build, and codegen drift check all clean; python -m backend.evals exits 0 with 4 passed / 1 not_implemented. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>
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
Validation
python -m compileall -q graphlink_apppython -m pytest -q --disable-warnings --maxfail=1— 428 passedThe local proposal remains intentionally excluded from the public repository under ignored
doc/.