Docker files restored for use on other machines with Docker/WSL2.
Available test methods:
1. Docker (isolated environment):
docker-compose -f docker/evolution-test/docker-compose.yml up evolution-feature
2. Local (bun runtime):
docker/evolution-test/run-local-test.bat feature
./docker/evolution-test/run-local-test.sh feature
Both methods provide:
- Millisecond precision timing
- Fitness score with 2 decimal places
- JSONL logging to .kilo/logs/fitness-history.jsonl
- Move docker-compose.evolution.yml to agent-evolution/docker-compose.yml
- Update README with current agent lineup (28+ agents)
- Fix model references in README tables
- Add recent commits history
- Simplify architecture overview
- Create agent-evolution/ directory with standalone dashboard
- Add interactive HTML dashboard with agent/model matrix
- Add heatmap view for agent-model compatibility scores
- Add recommendations tab with optimization suggestions
- Add Gitea integration preparation (history timeline)
- Add Docker configuration for deployment
- Add build scripts for standalone HTML generation
- Add sync scripts for agent data synchronization
- Add milestone and issues documentation
- Add skills and rules for evolution sync
- Update AGENTS.md with dashboard documentation
- Update package.json with evolution scripts
Features:
- 28 agents with model assignments and fit scores
- 8 models with benchmarks (SWE-bench, RULER, Terminal)
- 11 recommendations for model optimization
- History timeline with agent changes
- Interactive modal windows for model details
- Filter and search functionality
- Russian language interface
- Works offline (file://) with embedded data
Docker:
- Dockerfile for standalone deployment
- docker-compose.evolution.yml
- docker-run.sh/docker-run.bat scripts
NPM scripts:
- sync:evolution - sync and build dashboard
- evolution:open - open in browser
- evolution:dashboard - start dev server
Status: PAUSED - foundation complete, Gitea integration pending
Updated documentation to reflect official JSON Schema:
- model: global default model
- small_model: small model for titles/subtasks
- default_agent: default agent (must be primary mode)
- skills.urls: URLs to fetch skills from
According to official JSON Schema:
- model (not defaultModel) - global default model
- small_model (not smallModel) - small model for titles
- default_agent (not defaultAgent) - default agent to use
Also added mode: primary for user-facing agents.
- code: ollama-cloud/qwen3-coder:480b (coding tasks)
- ask: qwen/qwen3.6-plus:free (codebase questions)
- plan: ollama-cloud/nemotron-3-super (task planning)
- debug: ollama-cloud/gpt-oss:20b (bug diagnostics)
This fixes the issue where default models were not set in Settings.
These agents are invoked by other agents (orchestrator/evaluator), not directly by user:
- agent-architect: invoked by capability-analyst
- browser-automation: invoked by orchestrator for E2E testing
- history-miner: invoked by orchestrator during [planned] phase
- product-owner: invoked by evaluator for process improvements
- prompt-optimizer: invoked by evaluator when score < 7
- system-analyst: invoked by orchestrator during [researching] phase
- visual-tester: invoked by orchestrator for visual regression
Mode 'all' should be used only for agents that can be both
primary (user-facing) and subagent (invoked by other agents).
- Add nemotron-3-nano:30b to Model Format table in KILO_SPEC.md
- Update Pipeline Agents table to match current agent definitions
- Restore sync-agents.cjs script (uses .kilo/agents/*.md as source of truth)
- Script does NOT require kilo-meta.json in root (keeps Kilo Code working)
Models in use:
- ollama-cloud/nemotron-3-super (9 agents) ✓ available
- ollama-cloud/glm-5 (4 agents) ✓ available
- ollama-cloud/qwen3-coder:480b (3 agents) ✓ available
- qwen/qwen3.6-plus:free (2 agents) ✓ available
- ollama-cloud/minimax-m2.5 (2 agents) ✓ available
- ollama-cloud/gpt-oss:120b (2 agents) ✓ available
- ollama-cloud/nemotron-3-nano:30b (1 agent) ✓ available (added)
- Update agent-architect to nemotron-3-super for detailed specs
- Update capability-analyst to nemotron-3-super for deeper analysis
- Update history-miner to nemotron-3-super for better relevance
- Update prompt-optimizer to qwen3.6-plus via OpenRouter
- Fix system-analyst model path (remove openrouter prefix)
Testing showed ~28% quality improvement across cognitive agents
- Update security-auditor to ollama-cloud/nemotron-3-super-120b-a12b
- Update performance-engineer to ollama-cloud/nemotron-3-super-120b-a12b
- Update memory-manager to ollama-cloud/nemotron-3-super-120b-a12b
- Update evaluator to ollama-cloud/nemotron-3-super-120b-a12b
- Update planner to ollama-cloud/nemotron-3-super-120b-a12b
- Update reflector to ollama-cloud/nemotron-3-super-120b-a12b
- Update system-analyst to ollama-cloud/glm-5
- Update go-developer to ollama-cloud/qwen3-coder:480b
- Update markdown-validator to ollama-cloud/nemotron-3-nano
- Update prompt-optimizer to ollama-cloud/nemotron-3-super-120b-a12b
- Update product-owner to ollama-cloud/glm-5
Based on archive/agent-model-recommendations.json analysis
- Add go-developer agent for Go backend development
- Add 8 Go skills: web-patterns, middleware, db-patterns,
error-handling, security, testing, concurrency, modules
- Add go.md rules file
- Update capability-index.yaml with Go capabilities
- Complete backend coverage for both NodeJS and Go
- Added planner, reflector, memory-manager to capability-index.yaml
- Extended capability_routing with cognitive capabilities
- Updated orchestrator.md with new agent permissions and routing table
- All 3 new agents integrated into workflow
Related: Issue #25 (Research Milestone)
- Add upload_screenshot function to gitea-commenting skill
- Support multipart/form-data for attachments
- Create .test/e2e_test_screenshots.py test script
- Support SVG screenshots as placeholders
- 2 attachments uploaded to Issue #12 successfully
Gitea API:
- POST /repos/{owner}/{repo}/issues/{id}/assets - Upload attachment
- POST /repos/{owner}/{repo}/issues/{id}/comments - Comment with image reference
- GET /repos/{owner}/{repo}/issues/{id}/assets - List attachments
Milestone #44: All issues completed
- Create Dockerfile.playwright with headed mode by default
- Create docker-compose.yml with MCP server, headed, and test profiles
- Update Playwright skill to recommend headed mode for observation
- Remove --headless flag so browser window is visible
- Add BROWSER_VISIBILITY.md guide for X11/Docker setup
- Add README.Docker.md with quick start instructions
Configuration:
- PLAYWRIGHT_MCP_HEADLESS=false (browser visible)
- Requires X11 display for Docker (DISPLAY=:0)
- Three profiles: default, debug, test
Refs: #12 in Milestone #44
