| name | GNO |
|---|---|
| last_updated | 2026-08-06 |
| generator | flow-next-strategy |
People doing knowledge-heavy work keep important context across notes, code, PDFs, Office files, mail, calendars, and transcripts. Finding a plausible passage is easy; reliably recovering the right evidence, respecting its privacy boundary, and proving what supported a conclusion is still fragmented, repetitive, and hard to trust.
GNO wins by making one local-first knowledge engine the inspectable evidence layer across human and agent workflows. Retrieval must prove readiness and provenance, derived context must remain bounded and reproducible, and every network boundary must be explicit rather than hidden behind convenience.
Primary: Privacy- and evidence-conscious builders and knowledge workers — they're hiring GNO to turn heterogeneous local files into trustworthy, reusable context for themselves and their AI agents without surrendering control of the source material.
- Task-grounding accuracy — share of pinned benchmark tasks completed correctly from GNO-provided evidence; measured in committed retrieval and agentic eval artifacts.
- Evidence efficiency — retrieval calls and model-visible context required per correctly completed task; measured by paired Context Capsule benchmarks.
- Claim support integrity — share of substantive generated claims linked to exact supporting spans, with unsupported drafts withheld; measured by verified-answer evals and receipts.
- Retrieval-proven activation — share of setup attempts that finish with a real corpus-derived lexical hit and a healthy, independently reported semantic state; measured by setup/status/doctor contract tests and release smokes.
- Cross-surface contract parity — supported capabilities that behave consistently across CLI, Web UI, REST, SDK, MCP, and daemon surfaces; measured by contract tests and packed-package smokes.
Improve retrieval quality, diagnosis, Context Capsules, verified answers, and replayable evaluation evidence.
Why it serves the approach: Trust comes from inspectable support and reproducible quality, not confident output.
Strengthen ingestion, indexing, capture, editing, graph relationships, change tracking, and recovery across heterogeneous local sources.
Why it serves the approach: One dependable local knowledge layer removes the fragmentation at the center of the target problem.
Keep CLI, Web UI, REST, SDK, MCP, and the resident runtime aligned on one index and one set of contracts.
Why it serves the approach: Reusable context only compounds when every workflow sees the same capabilities and truth.
Make evidence bundles, exports, and publishing portable while preserving provenance, explicit egress policy, and caller ownership.
Why it serves the approach: Local-first should enable deliberate sharing without turning into opaque sync or accidental data movement.