feat(190B): Coding Benchmark Panel — Wave 2 + Wave 3 (3-tab /benchmark UI with Chart.js)

[Copilot]

Adds full coding benchmark frontend on top of the Wave 1 backend. /benchmark gains two new tabs — Code Models (start/status) and Code Results (history + jobs table + Chart.js charts) — alongside the existing LLM Models tab. Also cleans up the coding-benchmark-* naming duplication left from earlier work.

Summary

Wave 2 + Wave 3 frontend for /benchmark. Introduces tab-switched UI for coding benchmark: configurator + live console (Wave 2), and history + jobs table + Chart.js visualizations (Wave 3). Consolidates component naming to benchmark-coding-* convention and removes duplicate files.

Scope

In scope:

• benchmark-coding-configurator.tsx — Ollama model × task multi-select, loop task, timeout, max_rounds, stop-on-failure
• benchmark-coding-console.tsx — live log stream + progress bar (completed/total jobs)
• benchmark-coding-results.tsx — run history list + jobs table + delete/clear; fully i18n-d
• benchmark-coding-charts.tsx — Chart.js canvas charts: pass-rate bar (per model) + timing stacked bar (warmup/coding/request); exports computePassRates + computeTimings as pure testable helpers
• use-coding-benchmark.ts — polling, start/status, delete/clearAll; exports resolvePollStatus + buildProgressLog for unit tests
• app/benchmark/page.tsx — 3 tabs: llmModels / codeModels / codeResults
• i18n en/pl/de: benchmark.tabs.{llmModels,codeModels,codeResults} + benchmark.coding.{config,console,results,charts}.* (zero hardcoded user-facing strings)
• tests/use-coding-benchmark.test.ts — 18 unit tests covering hook helpers + chart data adapters
• Backend tests: +3 tests for security-hardened _run_dir and request_wall_seconds field → 13/13 new lines = 100% coverage

Out of scope:

• Backend endpoint/schema changes
• OpenAPI codegen update (hand-written types in types.ts; regeneration deferred)

Deleted duplicates:

• coding-benchmark-configurator.tsx
• coding-benchmark-results.tsx

Quality Gates (Hard Gate)

• make pr-fast passed
• CI required checks passed (Forbidden Paths Guard, Architecture drift guard, Backend lite (pytest), Frontend lite (lint), OpenAPI Contract (export + TS codegen), SonarCloud Scan, Quick validator (syntax + CI-lite deps))
• For new/renamed tests: make test-catalog-sync + make test-groups-sync executed

Validation Report

Commands run

node --import tsx --test tests/use-coding-benchmark.test.ts
node --import tsx --test tests/*.test.ts
npx tsc --noEmit
make test-groups-check
.venv/bin/python -m pytest tests/test_coding_run_service.py tests/test_coding_run_routes.py \
  --cov=venom_core --cov-report=xml:test-results/sonar/python-coverage.xml -q
make check-new-code-coverage-diagnostics
make pr-fast

Results (pass/fail)

Command	Result
tests/use-coding-benchmark.test.ts	PASS — 18/18
tests/*.test.ts (web-next)	PASS — 187/187
tsc --noEmit (changed files)	PASS — 0 new errors
make test-groups-check	PASS
backend tests (42 total)	PASS — 42/42
make check-new-code-coverage-diagnostics	PASS
make pr-fast	PASS ✅

Changed-lines coverage

13/13 = 100.0% (threshold 80%)

New backend tests cover: _run_dir invalid UUID guard (line 217), path-traversal relative_to check (lines 222–223), and request_wall_seconds artifact field (line 155).

Risks / Limitations / Skips

• Chart.js via canvas, no react-chartjs-2: Dynamic import("chart.js/auto") inside useEffect — avoids SSR, consistent with existing chart.js dep only.
• Code Results charts show current run only: Charts render codingRun.jobs from polling state. History entries don't carry full job arrays; full per-run chart drill-down is a future enhancement.
• first_sieve_task hardcoded to "" in configurator — wiring exposed in type but not in UI; Wave 4 item.
• OpenAPI types not regenerated: coding schemas are hand-written in types.ts and will drift until make openapi-export && make openapi-codegen-types runs (CI gate catches this).

Evidence

# web-next unit tests
# tests 187 / suites 52 / pass 187 / fail 0

# backend new-code coverage
Changed lines coverage: 13/13 = 100.0% (required: 80.0%)
PASS: changed-lines coverage meets the threshold.

# hard gate
✅ PR fast check passed.

Warning

Firewall rules blocked me from connecting to one or more addresses (expand for details)

I tried to connect to the following addresses, but was blocked by firewall rules:

• huggingface.co
  • Triggering command: /home/REDACTED/work/Venom/Venom/.venv/bin/python /home/REDACTED/work/Venom/Venom/.venv/bin/python -u -c import sys;exec(eval(sys.stdin.readline())) (dns block)
• unreachable.invalid
  • Triggering command: /home/REDACTED/work/Venom/Venom/.venv/bin/python /home/REDACTED/work/Venom/Venom/.venv/bin/python -u -c import sys;exec(eval(sys.stdin.readline())) (dns block)

If you need me to access, download, or install something from one of these locations, you can either:

• Configure Actions setup steps to set up my environment, which run before the firewall is enabled
• Add the appropriate URLs or hosts to the custom allowlist in this repository's Copilot coding agent settings (admins only)

Original prompt

---

This section details on the original issue you should resolve

<issue_title>190B — Integracja benchmarków codingowych z panelem web-next (/benchmark)</issue_title>
<issue_description>## 190B — Integracja benchmarków codingowych z panelem web-next (/benchmark)

Cel PR 190B

Udostępnić użytkownikowi w UI samodzielne uruchamianie benchmarków codingowych (sanity/simple/complex/feedback-loop) bez użycia CLI.

Zakres runtime (jawnie)

1. Ten etap (190B) dotyczy benchmarków Ollama coding.
2. Testowane modele pochodzą z lokalnego stosu Ollama (/api/tags, ollama list).
3. Parametry i metryki (warmup/coding/request) są liczone dla wywołań /api/generate w Ollama.

Stan obecny (as-is)

1. Ekran /benchmark obsługuje benchmark modeli LLM (latencja/szybkość/VRAM).
2. Backend ma endpointy benchmarkowe dla obecnego flow (/api/v1/benchmark/*).
3. Skrypty coding benchmark istnieją (scripts/ollama_bench/*), ale nie są podpięte pod panel.

Zakres docelowy (to-be) — 3 zakładki

1. Zakładka 1: LLM Models (istniejąca, bez regresji).
2. Zakładka 2: Code Models (nowa):
   • wybór modeli (multi-select),
   • wybór zadań (python_sanity, python_simple, python_complex, python_complex_bugfix),
   • parametry uruchomienia: timeout, model-timeout-overrides, max-rounds, options (temperature/top_p), first-sieve-task, stop-on-failure,
   • tryb uruchomienia: pojedynczy task lub scheduler.
3. Zakładka 3: Code Results (nowa):
   • tabela wyników (PASS/FAIL, error, task, model),
   • metryki czasu: warmup_seconds, coding_seconds, request_wall_seconds, total_seconds,
   • wykresy porównawcze (ranking, czasy, pass rate),
   • historia uruchomień i możliwość odświeżenia/usunięcia wpisu.

Backend/API do dodania

1. Nowy namespace API, np.:
   • POST /api/v1/benchmark/coding/start
   • GET /api/v1/benchmark/coding/{run_id}/status
   • GET /api/v1/benchmark/coding/list?limit=...
   • DELETE /api/v1/benchmark/coding/{run_id}
   • DELETE /api/v1/benchmark/coding/all
2. Endpoint start uruchamia scheduler/runner i zwraca run_id.
3. Endpoint status zwraca:
   • stan (pending/running/completed/failed),
   • postęp jobów,
   • wyniki cząstkowe i finalne,
   • ścieżki artefaktów i metryki timing.
4. Kontrakt payload oparty o istniejące parametry skryptów scripts/ollama_bench/scheduler.py.

Frontend (web-next) — zadania implementacyjne

1. Rozszerzyć app/benchmark/page.tsx o tabs + osobne sekcje per zakładka.
2. Dodać hook use-coding-benchmark analogiczny do use-benchmark.
3. Dodać komponenty:
   • benchmark-coding-configurator.tsx
   • benchmark-coding-console.tsx
   • benchmark-coding-results.tsx
   • benchmark-coding-charts.tsx
4. Dodać typy w web-next/lib/types.ts.
5. Dodać tłumaczenia pl/en/de dla nowych etykiet i komunikatów.
6. Dodać testy UI/hooków dla nowego flow.

Wykresy (JQ/JS) — decyzja technologiczna

1. W projekcie już jest chart.js (web-next/package.json) i istnieje użycie (components/cockpit/token-chart.tsx).
2. Rekomendacja na PR 190B:
   • użyć istniejącego chart.js (bez dokładania nowej biblioteki),
   • przygotować lekką warstwę adaptera do danych benchmarkowych.
3. Jeśli potrzebna szybsza ergonomia komponentów:
   • opcjonalnie dodać react-chartjs-2 (wrapper do obecnego Chart.js).
4. jq (CLI) może służyć do pre-agregacji artefaktów JSON po stronie dev/ops, ale nie jest biblioteką do renderu wykresów w UI.

KPI i prezentacja na zakładce Code Results

1. pass_rate per model.
2. Mediany czasów:
   • warmup,
   • coding,
   • request.
3. Liczba fail z podziałem na kategorię:
   • timeout,
   • empty response,
   • invalid format,
   • test/lint failure.
4. Wykresy:
   • bar chart: pass_rate per model,
   • stacked bar: warmup vs coding,
   • table + sort + filters (task/model/status/date).

DoD dla PR 190B

1. /benchmark ma 3 działające zakładki (LLM Models, Code Models, Code Results).
2. Użytkownik może uruchomić coding benchmark z UI i śledzić status.
3. Wyniki i metryki timing są prezentowane tabelarycznie i graficznie.
4. Historia uruchomień jest dostępna i odświeżalna.
5. Brak regresji istniejącej zakładki LLM benchmark.
6. i18n parity: wszystkie nowe teksty w pl/en/de.

Plan realizacji falami (mniejszy zakres na sesję agenta)

1. Fala 1: backend API + kontrakty danych
   • Zakres:
     • endpointy /api/v1/benchmark/coding/*,
     • serwis uruchamiania benchmarku i statusów,
     • schematy request/response,
     • wiring w main.py i router_wiring.py,
     • testy backendowe (service/routes).
   • Poza zakresem:
     • zmiany UI zakładek /benchmark,
     • wykresy frontend.
   • DoD fali:
     • endpointy działają lokalnie,
     • testy backendowe zielone,
     • brak regresji istniejącego API benchmark.
2. Fala 2: frontend Code Models (uruchamianie testów)
   • Zakres:
     • 2. zakładka Code Models,
     • use-coding-benchmark,
     • konfigurator + k...

• Fixes 190B — Integracja benchmarków codingowych z panelem web-next (/benchmark) #427

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.

[Copilot]

Pull request overview

Adds the Wave 2/3 “coding benchmark” frontend to /benchmark (3-tab UI) and introduces a new backend coding-benchmark service + API namespace to start/poll/list/delete runs, including persisted run metadata and job timing enrichment from artifacts.

Changes:

• Frontend: new /benchmark tabs for Code Models (run config + live console) and Code Results (history + jobs table + Chart.js charts) plus a polling hook and TS types.
• Backend: new CodingBenchmarkService that launches the existing scheduler script, persists run state, enriches job timing from artifacts, and exposes /api/v1/benchmark/coding/* routes + schemas.
• Testing/Docs: adds unit/contract tests, registers them in test catalog/groups, and updates agent docs/templates for new-test naming/checklist.

Reviewed changes

Copilot reviewed 25 out of 25 changed files in this pull request and generated 18 comments.

Show a summary per file

File	Description
web-next/app/benchmark/page.tsx	Adds tabbed /benchmark UI and wires coding benchmark components/hook.
web-next/hooks/use-coding-benchmark.ts	New polling + start/delete/clear hook (and exported helpers).
web-next/components/benchmark/benchmark-coding-configurator.tsx	UI to select Ollama models/tasks and run parameters.
web-next/components/benchmark/benchmark-coding-console.tsx	Live log viewer + progress bar for coding benchmark runs.
web-next/components/benchmark/benchmark-coding-results.tsx	History list + current run jobs table + delete/clear actions.
web-next/components/benchmark/benchmark-coding-charts.tsx	Chart.js charts + pure adapters for pass rates/timings.
web-next/lib/types.ts	Adds coding benchmark request/response/job/run types for frontend.
web-next/lib/i18n/locales/{en,pl,de}.ts	Adds translation keys for new benchmark tabs/coding sections.
web-next/tests/use-coding-benchmark.test.ts	Adds unit tests for hook/chart pure helpers.
venom_core/services/benchmark_coding.py	Implements coding benchmark run lifecycle, persistence, scheduler launch, artifact enrichment.
venom_core/api/schemas/benchmark_coding.py	Pydantic schemas for coding benchmark endpoints.
venom_core/api/routes/benchmark_coding.py	FastAPI routes for start/status/list/delete/clear-all coding benchmarks.
venom_core/main.py	Initializes and injects CodingBenchmarkService; registers router.
venom_core/bootstrap/router_wiring.py	Wires coding benchmark router dependencies.
tests/test_coding_run_service.py	Unit tests for CodingBenchmarkService (incl. security guards + artifact timing field).
tests/test_coding_run_routes.py	Contract tests for coding benchmark API router.
docs/{AGENTS.md,PL/AGENTS.md}	Documents required new-test naming/checklist + i18n rule.
config/testing/test_catalog.json	Registers new backend tests in the catalog.
config/pytest-groups/{sonar-new-code,fast,ci-lite}.txt	Adds the new backend tests to lanes/groups.
.github/coding-agent-task-template.md	Updates template with naming/checklist guidance for new tests.

[mpieniak01]

Uwagi z przeglądu #430 (review)
uwzględnione w
1f8e8b7