# Changelog of Super AI Bench (MCP) (`flamboyant_leaf/super-ai-bench-mcp`) Actor

- **URL**: https://apify.com/flamboyant\_leaf/super-ai-bench-mcp/changelog.md
- **Full Actor documentation**: https://apify.com/flamboyant\_leaf/super-ai-bench-mcp.md

## Changelog

### 2026-01-20

- Added `official` boolean filter parameter to `search_replicate_models_bulk` tool. When set to `true`, filters results to only include official models (default: `false`).
- Added `is_official` flag to Replicate search results. Models are enriched with official status from Replicate's `/collections/official` endpoint (cached for 24 hours).
- Removed `cover_image_url` and `default_example` fields from Replicate search responses to reduce response size.
- Removed `jq` parameter from `search_replicate_models_bulk` tool - now returns cleaned responses directly.

### 2026-01-17

- Added jq preview tools for LLM and text-to-image benchmarks; benchmark errors now return preview hints (sample + fields).
- Removed image-editing, speech, and video benchmark tools/endpoints to focus on text and text-to-image only.
- Added `get_readme_documentation` tool as an alias of `get_protocol_documentation`.
- Increased Replicate bulk search max to 10 via `AA_REPLICATE_SEARCH_MAX` (predictions remain capped by `AA_REPLICATE_BULK_MAX`).
- Updated documentation (README + protocol prompt) to reflect the current toolset, jq/preview workflow, and timeout guidance.