v0.0.4

v0.0.4 Release

Optimized API directory auto-refresh mechanism

• Shortened the default cache refresh interval to 1 hour (from 6 hours)

• Added forced refresh function, supporting immediate retrieval of the latest APIs

• Added directory refresh logs for easier troubleshooting

• Updated documentation

Official documentation address: https://video-ai.apifox.cn

Full Changelog: v0.0.3...v0.0.4

v0.0.3

v0.0.3

This release adds runtime auto-refresh support for the official Apifox API catalog, making LittleOrange Video MCP easier to keep in sync with newly released LittleOrange video generation APIs.

Highlights

- Added runtime auto-refresh for the Apifox API catalog.
- New official API interfaces can be discovered without requiring a new PyPI/package release.
- Improved MCP configuration and documentation around automatic updates.
- Continued support for running the MCP server via:

bash
uvx littleorange-video-mcp


Changes since v0.0.2

- feat: auto-refresh Apifox catalog at runtime
- Enable auto-update for MCP
- README updates for improved clarity and new features
- Resolve README merge conflict

This allows the MCP wrapper to refresh from the official Apifox documentation so newly added LittleOrange video APIs can become available without waiting for another package release.

v0.0.2

Release Notes

Highlights

This release packages the first complete external-facing version of the LittleOrange Video MCP usability improvements into one version.

Key themes:

• better configurability
• better polling stability
• better agent/tool usability
• better diagnostics and error reporting
• better TRAE/uvx onboarding

Added

Configuration

• Added LITTLEORANGE_BASE_URL
• Added LITTLEORANGE_POLL_INTERVAL_SECONDS
• Added LITTLEORANGE_MAX_POLL_ATTEMPTS
• Added LITTLEORANGE_FIRST_POLL_DELAY_SECONDS
• Added LITTLEORANGE_DEBUG
• Added LITTLEORANGE_LOG_FILE
• Added per-call overrides for:
  • base_url
  • poll_interval_seconds
  • max_poll_attempts
  • first_poll_delay_seconds

Polling improvements

• Added first-poll delay support for async task creation flows
• Added normalized polling state handling
• Added richer polling result fields:
  • elapsed_seconds
  • last_state
  • last_status
  • last_error
  • normalized_status
• Improved query argument propagation from create flow to query flow

Agent-friendly tools

Added high-level tools for better IDE/agent ergonomics:

• video_generate_wait
• image_to_video_wait
• video_extend_wait
• video_query
• asset_upload
• asset_list

Raw passthrough improvements

• Added headers
• Added query_params
• Added support for arbitrary JSON request_body

Diagnostics and errors

• Added structured JSON error payloads
• Improved HTTP error context with safe redaction
• Added config validation with user-friendly messages
• Added optional file-based debug logging

Docs

• Expanded README substantially
• Added TRAE-ready configuration example
• Added recommended polling presets
• Added FAQ and debug/logging guidance

Changed

• Centralized config handling in littleorange_video_mcp/config.py
• Improved tool descriptions to steer users/agents toward _wait tools when appropriate
• Kept tool names under the TRAE 60-character constraint

Tests

• Added coverage for config parsing and validation
• Added coverage for schema exposure of new fields
• Added coverage for polling normalization and timeout/failure handling
• Added coverage for raw request enhancements
• Retained tool-name-length guard tests

Compatibility

• No intended breaking changes for existing basic tool usage
• Existing direct API-style tools remain available
• New features are additive, with improved structured error output

Version

• Bumped package version from 0.0.1.post2 to 0.0.2

v0.0.1-rolo

Dreamina-Seedance 2.0 MCP compatibility release.

• Fixes Dreamina text content schema so text items do not require or send role.
• Keeps media content roles such as reference_image, first_frame, last_frame, reference_video, and reference_audio.
• Loosens Dreamina create-video required body fields to match official examples.
• PyPI package version: 0.0.1.post1.

v0.0.1

v0.0.1

Initial private release of littleorange-video-mcp.

Highlights

• Unified package name: littleorange-video-mcp
• Python import package: littleorange_video_mcp
• MCP server name: littleorange-video-mcp
• 36 documented LittleOrange video API tools
• 11 auto-poll _wait tools that create tasks and return final video URLs when available
• Raw passthrough tool: littleorange_raw_request
• Environment variables: LITTLEORANGE_API_KEY, LITTLEORANGE_TIMEOUT
• Public docs sanitized to avoid local build paths

MCP config after PyPI publish

{
  "mcpServers": {
    "littleorange-video": {
      "command": "uvx",
      "args": ["littleorange-video-mcp"],
      "env": {
        "LITTLEORANGE_API_KEY": "sk-你的key",
        "LITTLEORANGE_TIMEOUT": "120"
      }
    }
  }
}

Before PyPI publish, use the GitHub tag source with uvx --from git+https://github.com/littleorange-ai/littleorange-video-mcp.git@v0.0.1 littleorange-video-mcp.

0.0.1.post2

Summary

• Tolerate MCP clients that inject request_body: null into no-body query tools.
• Fix raw passthrough argument filtering so method/path/base_url do not fail strict validation.
• Bump package version to 0.0.1.post2.

Verification

• python3 -m unittest discover -s tests -v
• python3 -m py_compile littleorange_video_mcp/*.py run_server.py
• JSON schema serialization for 36 operations
• uv build
• uvx twine check dist/*
• Wheel install/request construction smoke test