ckvd-research（CVD 研究）

CKVD Codebase Research

Research the crypto-kline-vision-data codebase to answer questions about: $ARGUMENTS

Research Focus

Key Areas to Investigate

1. FCP Implementation (src/ckvd/core/sync/crypto_kline_vision_data.py)

- How failover decisions are made - Cache → Vision → REST priority

1. Provider Implementations (src/ckvd/core/providers/binance/)

- Vision API client - REST API client - Cache manager

1. Utilities (src/ckvd/utils/)

- Market constraints and validation - Timestamp handling - DataFrame utilities

1. Streaming Implementation (src/ckvd/core/streaming/)

- KlineStream class and stream lifecycle - KlineUpdate dataclass structure - StreamConfig configuration options - WebSocket client connection handling - Async message loop patterns

Research Instructions

1. Use Glob to find relevant files
2. Use Grep to search for specific patterns
3. Read key files to understand implementation
4. Summarize findings with specific file:line references

Expected Output

Provide:

• Summary: Brief answer to the research question
• Key Files: Files most relevant to the topic
• Code References: Specific file:line references
• Related Topics: Other areas worth investigating

TodoWrite Task Templates

Template A: Investigate FCP Flow

1. Read src/ckvd/core/sync/crypto_kline_vision_data.py get_data() method
2. Trace cache check logic (get_cache_lazyframes in ckvd_cache_utils)
3. Trace Vision fetch logic (_fetch_from_vision)
4. Trace REST fallback logic (_fetch_from_rest)
5. Document FCP decision points with file:line references

Template B: Trace Data Source

1. Identify the data source module (Vision, REST, Cache)
2. Read the provider client implementation
3. Trace data flow from API call to DataFrame return
4. Document error handling and retry logic
5. Summarize with file:line references

Template C: Map Exception Handling

1. Read rest_exceptions.py and vision_exceptions.py
2. Grep for exception catch/raise patterns in core/
3. Map which exceptions trigger FCP fallback
4. Identify any silent failures or bare excepts
5. Document exception flow with file:line references

Template D: Investigate Streaming Architecture

1. Read src/ckvd/core/streaming/kline_stream.py (KlineStream class)
2. Understand stream lifecycle: create → subscribe → iterate → close
3. Research KlineUpdate dataclass (fields: open_time, close, open, high, low, volume, symbol, interval, is_closed)
4. Trace async context manager patterns (__aenter__, __aexit__)
5. Study WebSocket client initialization and message handling
6. Document stream state management and error handling
7. Summarize async patterns with file:line references

Post-Change Checklist

After modifying this skill:

• Key areas to investigate still match actual file paths
• Research instructions reference available tools
• Append changes to references/evolution-log.md

References

• @references/evolution-log.md - Skill change history