Skip to content

Latest commit

 

History

History
170 lines (128 loc) · 5.78 KB

File metadata and controls

170 lines (128 loc) · 5.78 KB

CV181X/CV182X/CV180X Multimedia API Skill

Maintained by: Seeed Studio
Repository: https://github.com/Seeed-Studio/ai-skills
Original work by: Pillar1989

Expert guidance skill for CV181X/CV182X/CV180X multimedia development on Sophgo platforms (reCamera).

Overview

This skill provides comprehensive knowledge of:

  • VI (Video Input)
  • VPSS (Video Processing)
  • VENC (Video Encoding)
  • VDEC (Video Decoding)
  • VO (Video Output)
  • Audio (AI/AO/AENC/ADEC/VQE)
  • SYS (System Control)
  • VB (Video Buffer Pool)
  • REGION (Regional Management/OSD)
  • GDC (Geometric Distortion Correction Subsystem)
  • Debugging and troubleshooting

Version

Current Version: v2.3.0

What's New in v2.3.0

  • Overview + Platform References: Added overview.md and platform.md for system workflow, SDK layout, and pixel formats
  • Integration Guide: Added integration-guide.md for cross-module design and triage
  • Binding Cookbook Consolidation: Concurrent scenarios fully consolidated into binding-cookbook.md
  • Binding Matrix Clarified: VDEC -> VPSS bind vs SendFrame options clarified and aligned with SDK samples
  • VPSS Limits Aligned: Scaling limits and channel counts aligned with SDK defines

What's New in v2.1.0

  • Source Code Verification: All APIs verified against cvi_mpi source code
  • Critical Fixes: VENC SendFrame memory requirements corrected (must use VB Pool)
  • Missing APIs Added: VENC SendFrameEx, USER_FRAME_INFO_S, GDC CancelJob, GDC MESH structures
  • Module List Complete: Added H265D, H264D, VPSSF, TPU to MOD_ID_E enumeration
  • ION Cache Management: Added IonFlushCache/IonInvalidateCache documentation
  • VPSS Constraints: Documented input source mode restrictions (Bind vs SendFrame)
  • Binding Cookbook: Consolidated concurrent scenarios with TPU inference path
  • VB Pool Enhancements: VB_INVALID_POOLID clarification and EX mode documentation

See CHANGELOG.md for complete details.

Structure

cv181x-media/
├── SKILL.md              # Main skill definition
├── references/           # Module reference documentation
│   ├── vi.md            # Video Input (4-layer architecture)
│   ├── vpss.md          # Video Processing (10 features)
│   ├── venc.md          # Video Encoding
│   ├── vdec.md          # Video Decoding (NEW)
│   ├── vo.md            # Video Output
│   ├── audio.md         # Audio Subsystem (NEW)
│   ├── sys.md           # System Control (11 categories)
│   ├── vb.md            # Video Buffer (8 categories)
│   ├── rgn.md           # Region Management
│   ├── gdc.md           # Geometric Distortion Correction
│   ├── debug.md         # Debugging guide
│   ├── troubleshooting.md
│   ├── overview.md      # System overview and workflow
│   ├── platform.md      # SDK layout and platform limits
│   ├── integration-guide.md # Cross-module integration rules
│   └── scenarios.md     # Common scenarios
├── scripts/              # Automation scripts
│   ├── update_from_sdk.sh
│   ├── learn_from_usage.py
│   └── validate_skill.sh
└── .skillrc              # Skill configuration

Self-Learning and Auto-Update

This skill is designed to continuously improve through:

1. SDK Updates

Automatically extract and integrate new API information from SDK releases:

./scripts/update_from_sdk.sh /path/to/new/sdk

2. Usage Learning

Learn from actual usage patterns and collect feedback:

./scripts/learn_from_usage.py --feedback-file usage_log.txt

3. Version Control

All changes are tracked via git:

git log --oneline  # View change history
git diff v1.0.0    # Compare with previous version

Update Workflow

  1. Detect SDK Update: Monitor SDK release notes
  2. Extract Changes: Parse new headers and documentation
  3. Update References: Regenerate affected reference docs
  4. Validate: Run validation checks
  5. Commit: Version control with semantic versioning
  6. Package: Create new .skill file

Usage

Install this skill in Claude Code or Codex:

# Claude
claude skills install git+https://github.com/Seeed-Studio/ai-skills#subdirectory=skills/cv181x-media

# Codex
codex skills install git+https://github.com/Seeed-Studio/ai-skills#subdirectory=skills/cv181x-media

Once installed, ask your AI assistant questions like:

  • "How to configure VI module to capture 1080p video?"
  • "Show me how to add timestamp OSD to video stream"
  • "How to correct barrel distortion from wide-angle lens?"

Contributing

To improve this skill, please contribute to the main repository: https://github.com/Seeed-Studio/ai-skills

  1. Fork the repository
  2. Create feature branch: git checkout -b feature/cv181x-media-improvement
  3. Make improvements
  4. Submit pull request

See CONTRIBUTING.md for detailed guidelines.

Feedback Collection

When using this skill, report issues or improvements:

  • API changes discovered
  • Missing information
  • Incorrect documentation
  • New use cases

Maintenance

Regular Tasks

  • Monthly: Check for SDK updates
  • Quarterly: Review usage patterns and update common scenarios
  • Annually: Major version release with comprehensive review

Changelog

See CHANGELOG.md for version history.

License

This skill is licensed under the MIT License.

Acknowledgments

This skill was originally created by Pillar1989 and is now maintained by Seeed Studio for the reCamera developer community.

Authors

  • Original creation: Pillar1989, 2026-01-17
  • Maintained by: Seeed Studio, 2026-01-27+
  • Auto-update system: Enabled