siftq
Sign inSearch API Playground
Sign inSearch API Playground
Changelog

What shipped.

Every public SiftQ release, oldest at the bottom.

LATESTv0.2.0· 2026-05-10· Multi-modal scopes (image / video / podcast) reach general availability.
Try the latest SiftQ API release →

v0.2.0

Multi-modal scopes (image / video / podcast) reach general availability.

Added
  • —image scope: returns imageUrl, imageWidth, imageHeight directly
  • —video scope: includes duration and coverImage fields
  • —podcast scope: duration in seconds for time-budgeted listening agents
  • —scholar scope for academic paper search across arXiv, DOI, and major preprint servers
Changed
  • —Default p50 latency cap lowered from 220ms to 180ms after edge rollout
  • —Free tier raised from 500 to 1,000 queries / month
Fixed
  • —Edge cases where document scope occasionally returned empty for valid PDF queries
  • —Concise snippet truncation could cut mid-sentence on rare Unicode boundaries

v0.1.5

EU region (Frankfurt) launched. Pro tier customers now get region-locked routing.

Added
  • —Region: eu-central-1 (Frankfurt) for sub-50ms p50 from European clients
  • —X-Siftq-Region response header for verifying which region served the request
  • —Per-region rate limit isolation

v0.1.4

MCP server v1 — drop-in support for Claude Desktop and any MCP-compatible client.

Added
  • —@siftq/mcp-server published on npm
  • —Server exposes search as a single tool with scope as a parameter
  • —Tool descriptions tuned so Claude auto-picks the right scope per question
Changed
  • —Bearer token auth now accepts both "Bearer mk-..." and bare "mk-..." for compatibility

v0.1.3

Structured output mode reaches GA.

Added
  • —includeSummary parameter: LLM-generated overview of top hits
  • —conciseSnippet parameter: trim to ~3 sentences of exact-match text
  • —OpenAPI 3.1 spec published at /openapi.yaml and /openapi.json
Fixed
  • —Authors array could contain duplicate names in academic scope

v0.1.2

Official LangChain and LlamaIndex SDKs.

Added
  • —langchain-siftq (Python): SiftQRetriever drop-in for any chain or agent
  • —llama-index-retrievers-siftq (Python): BaseRetriever subclass
  • —siftq-node and siftq-python: first-party SDKs with typed clients

v0.1.1

Highlight extraction and token-efficient snippets.

Added
  • —Per-result score field (low / medium / high) for downstream filtering
  • —authors[] for academic and book scopes
Changed
  • —Default snippet length lowered to reduce token bloat in RAG prompts

v0.1.0

Public preview launch.

Added
  • —POST /v1/search with webpage, news, academic, book, all, document scopes
  • —Free tier: 500 queries / month, no credit card
  • —JS / TS / Python / Go quickstart examples
Stay current

New release every 4-6 weeks.

Try the latest SiftQ API release →