Перейти к содержанию

Critical Audit — v2.0.0-alpha → v2.0.0 Stable

Date: 2026-07-03 Auditor: AI Agent (GLM-5.2) Status: Complete

1. Version Consistency

File Current Expected Status
src/lib/00-core.sh v2.0.0-alpha v2.0.0 OK (bump at release)
setup.sh:3 v1.1.0 v2.0.0 FIX
README.md:1 v1.1.0 v2.0.0 FIX
CHANGELOG.md links to v1.0.0/v1.1.0 add v2.0.0 FIX
AGENTS.md v2.0.0-alpha v2.0.0 OK (bump at release)

2. Module Numbering

Duplicates found: - 22-mise.sh + 22-webui-service.sh — both use prefix 22 - 32-isolated.sh + 32-observability.sh — both use prefix 32

Modules not in AGENTS.md: - 32-observability.sh — Prometheus + Grafana stack - 33-gui.sh — Web management interface

Fix: Renumber to avoid duplicates: - 22-webui-service.sh → keep as 22-webui-service.sh, rename 22-mise.sh → 29-mise.sh - 32-observability.sh → 34-observability.sh - 33-gui.sh → 35-gui.sh - 32-isolated.sh → keep as 32-isolated.sh

3. Provider Registry Gaps

Providers in opencode (from source code analysis)

opencode bundles 22 SDK providers: 1. @ai-sdk/anthropic — Anthropic Claude 2. @ai-sdk/amazon-bedrock — Amazon Bedrock 3. @ai-sdk/azure — Azure OpenAI 4. @ai-sdk/google — Google Gemini 5. @ai-sdk/google-vertex — Google Vertex AI 6. @ai-sdk/openai — OpenAI 7. @ai-sdk/openai-compatible — OpenAI-compatible (z.ai, Moonshot, etc.) 8. @openrouter/ai-sdk-provider — OpenRouter 9. @ai-sdk/xai — xAI Grok 10. @ai-sdk/mistral — Mistral 11. @ai-sdk/groq — Groq 12. @ai-sdk/deepinfra — DeepInfra 13. @ai-sdk/cerebras — Cerebras 14. @ai-sdk/cohere — Cohere 15. @ai-sdk/gateway — Gateway 16. @ai-sdk/togetherai — Together AI 17. @ai-sdk/perplexity — Perplexity 18. @ai-sdk/vercel — Vercel 19. @ai-sdk/alibaba — Alibaba/Qwen 20. gitlab-ai-provider — GitLab 21. @ai-sdk/github-copilot — GitHub Copilot 22. venice-ai-sdk-provider — Venice AI

Providers in project (26-providers.sh)

20 providers registered: 1. deepseek ✓ 2. opencode ✓ 3. xai ✓ 4. mimo ✓ 5. moonshot ✓ 6. minimax ✓ 7. openai ✓ 8. anthropic ✓ 9. google ✓ 10. mistral ✓ 11. groq ✓ 12. together ✓ 13. cohere ✓ 14. fireworks ✓ (not in opencode bundled — use openai-compatible) 15. cerebras ✓ 16. perplexity ✓ 17. ollama ✓ (local) 18. litellm ✓ (local) 19. vllm ✓ (local) 20. sglang ✓ (local)

Missing providers (CRITICAL)

Provider Priority Reason
z.ai (GLM) CRITICAL Key provider for RU/CN users. GLM-5.2 is current flagship. Uses openai-compatible SDK.
OpenRouter HIGH Aggregator access to 100+ models via single API key.
Alibaba/Qwen HIGH Major Chinese provider. Qwen3 series. Native SDK in opencode.
DeepInfra MEDIUM Fast inference, competitive pricing.
GitHub Copilot MEDIUM Free for GitHub users. Native SDK.
Venice AI LOW Privacy-focused. Niche.

Model version audit

Provider Current in project Actual latest (2026-07) Status
DeepSeek V4 Pro / V4 Flash V4 Pro / V4 Flash OK
xAI Grok 3 Grok 4 UPDATE
Moonshot Kimi K2.6 Kimi K2 VERIFY
Anthropic Claude 4 Claude 4 Opus / Sonnet 4.5 UPDATE
Google Gemini 2.5 Gemini 2.5 Pro OK
OpenAI GPT-5 GPT-5 OK
Mistral Large 3 Large 3 OK
z.ai — (missing) GLM-5.2 / GLM-4.6 ADD

4. Infrastructure

Missing files: - infra.yml — referenced in CHANGELOG but not found in repo - Docker Compose file for PostgreSQL + Qdrant + Redis + Prometheus + Grafana + MemoryLayer

Fix: Create infra.yml or verify it's generated by 30-infra.sh at runtime.

5. Test Coverage

Module Unit test Status
00-core.sh test_core.sh OK
helpers.sh test_helpers.sh OK
30-infra.sh test_infra.sh OK
31-cockpit.sh test_cockpit.sh OK
33-gui.sh test_gui.sh OK
12-mcp-lsp.sh test_mcp_registry.sh OK
18-opencode-json.sh test_opencode_json_gen.sh OK
32-isolated.sh — MISSING
32-observability.sh — MISSING
26-providers.sh — MISSING
embed-proxy.py — MISSING

6. CI/CD Gaps

Workflow Covers Missing
test.yml bash tests Go tests (cockpit), Python tests (embed-proxy)
shellcheck.yml shell scripts —
docs.yml docs build —

Fix: Add Go test step to test.yml, add Python syntax check for scripts/.

7. Documentation

Doc Status Action
README.md v1.1.0, outdated Rewrite for v2.0.0
CHANGELOG.md Missing v2.0.0 section Add v2.0.0 entry
AGENTS.md Mostly current Add missing modules, fix numbering
docs/ (landing) May be outdated Audit and update
CONTRIBUTING.md Needs check Verify accuracy

8. opencode.json

Current model: deepseek/deepseek-v4-pro Issue: No z.ai/GLM provider configured, despite being a key use case.

Fix: Add z.ai provider with GLM-5.2 as option, add OpenRouter as fallback aggregator.

Action Plan

Phase A: Critical Fixes (before Phase 3)

  1. Fix version consistency (setup.sh, README.md)
  2. Fix module numbering duplicates
  3. Add z.ai (GLM) provider to registry + opencode.json
  4. Add OpenRouter provider
  5. Update model versions (Grok 4, Claude 4 Opus/Sonnet)
  6. Create/verify infra.yml
  7. Update AGENTS.md with all modules

Phase B: Phase 3 — Tests & CI

  1. Add unit tests for 32-isolated.sh, 32-observability.sh, 26-providers.sh
  2. Add Python test for embed-proxy.py
  3. Add Go test step to CI
  4. Add Python syntax check to CI

Phase C: Finalization

  1. Rewrite README.md for v2.0.0
  2. Update CHANGELOG.md with v2.0.0
  3. Update docs/ landing page
  4. Bump version to v2.0.0
  5. Tag release