API preview

Search expert AI engineering clips programmatically.

Production v1 API with authority explainability, Stripe billing, and B2B workspaces. Retrieval responses expose why each source was chosen — not opaque vector scores.

What the API would expose

• Retrieval APIs over indexed transcript moments
• Expert clip search with confidence and quality signals
• Timestamp retrieval with YouTube deep links
• Engineering knowledge indexing by topic and comparison

Use cases we are validating

AI startups

Ground copilots in practitioner clips instead of generic blog summaries.

Coding assistants

Attach timestamp citations when explaining RAG, agents, or infra tradeoffs.

Internal copilots

Index long-form engineering talks your team already trusts.

Enterprise research systems

Feed retrieval eval loops with expert-sourced moments.

AI education tools

Surface the best 30-second explanation per concept from real talks.

Trust product layer for rag evaluation — cluster evals-observability

Why this answer is trustworthy

Trust 71%Enterprise readiness 90%Evidence strength 64%Diversity 100%

Why this answer won

Tier-1 expert moment (Anthropic) paired with hard doc citations; fusion=rrf and top_k=20 in operational excerpt.

Configs used

top_k=8
corpus · confidence 43%

Benchmark evidence

faithfulness=0.82
observed in cited evidence
Sam Witteveen

Failure fixes

No incident remediation path in cited evidence.

Expert video corroboration

Building with MCP and Claude API — tool invocation

Sam Witteveen

https://www.youtube.com/watch?v=aZLr962R6Ag&t=638

Contradictory evidence

No contradictory expert framing detected.

Trace lineage

queryretrieval.request
hybrid_search
rag evaluation
retrieve_hit_1retrieval.candidate
Sam Witteveen
12:34 · score 0.82
synthesisanswer.operational_gate
mcp_tool_orchestration
passed

Citation quality (primary)

RAG vs MCP — expert walkthrough

Authority 82%· high

MCP is a protocol for tool context, while RAG retrieves documents at answer time. trace id visible in LangSmith retrieve span; faithfulness=0.82.

Source type:: youtube_expert
Cluster:: evals-observability

Citation →Timestamp →

Authority 82% · high confidence

Winning evidence

expert Building with MCP and Claude API — tool invocation90%
citation RAG vs MCP — expert walkthrough82%
config top_k=843%
trace LangSmith32%
trace retrieve span32%

Rejected evidence

Excluded candidates: lower rank or diversity cap

Operational checklist

✓ Hard citations paired — 1 cited moment(s)
✓ Configuration evidence
✓ Benchmark / metric evidence
✓ Trace / observability lineage
✓ Failure / remediation evidence
✓ Expert video corroboration — Building with MCP and Claude API — tool invocation
✓ Source diversity — 100%
✓ Contradictions reviewed

Uncertainty

Static trust demo; run live /api/v1/answer for production trust envelope.