…ployment docs (#4)

- .env.example: Add Evolution Mode config with description
- deploy/aliyun/.env.prod.example: Add production Evolution Mode section
- docker/compose/prod.yml: Pass EVOLUTION_ENABLED and EVOLUTION_ADMIN_ONLY env vars
- docs/deployment/BINARY_DEPLOYMENT.md: Add Evolution Mode setup instructions

Evolution Mode is an advanced form of Geek Mode that unlocks full Claude Code Agent capabilities.

Co-authored-by: Test User <test@example.com>

Add ai.chat.user_message log to help diagnose issue where
messages are silently swallowed after previous block is done.

Related to #4

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

Root cause: When AmazingParrot's plan sets needs_direct_answer=true
(with no memo_search/schedule_query/etc), the retrievalResults map
is empty. This caused buildSynthesisPrompt to generate a synthesis
prompt with an empty context placeholder, which could cause the LLM
to hang or produce unexpected behavior.

Changes:
- Add fallback casual chat prompt when no retrieval results exist
- Add extensive debug logging to track synthesis flow
- Add chunk counting and channel close logging for better visibility

This fixes the bug where messages are "swallowed" after a block
completes with status=done.

Related: Task #4

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* fix(frontend): optimize Vite config for React deps stability

Add React and related dependencies to optimizeDeps.include to prevent
runtime issues where React hooks may be unavailable due to module
loading order problems.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* fix(ai): add debug log for user message event tracking

Add ai.chat.user_message log to help diagnose issue where
messages are silently swallowed after previous block is done.

Related to #4

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* fix(ai): resolve message hang when needs_direct_answer=true

Root cause: When AmazingParrot's plan sets needs_direct_answer=true
(with no memo_search/schedule_query/etc), the retrievalResults map
is empty. This caused buildSynthesisPrompt to generate a synthesis
prompt with an empty context placeholder, which could cause the LLM
to hang or produce unexpected behavior.

Changes:
- Add fallback casual chat prompt when no retrieval results exist
- Add extensive debug logging to track synthesis flow
- Add chunk counting and channel close logging for better visibility

This fixes the bug where messages are "swallowed" after a block
completes with status=done.

Related: Task #4

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* feat(ai): add progress events for chat response feedback

Add immediate feedback events to eliminate 5-10 second "dead silence"
after user sends message. Implements 200ms psychological threshold
principle for user experience.

Backend changes:
- Add "received" event sent immediately after message receipt
- Add "routing_start" event before intent classification
- Add "routing_end" event with agent info and duration
- Apply events consistently across Normal/Geek/Evolution modes

Refs #89

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* fix(tests): correct JSON field names in EventMeta test

The test was checking for PascalCase field names (DurationMs, ToolName)
but the struct uses snake_case JSON tags (duration_ms, tool_name).

This is a pre-existing bug in PR #65 that caused CI to fail.

Refs #89

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* style: gofmt format alignment fixes

Fix gofmt alignment issues detected by CI golangci-lint.

Refs #90

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

---------

Co-authored-by: 黄飞虹 <aaronwong1989@gmail.com>
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>