AURORA Agent (bioprism)

Query-compiled inference for executable biology.

Context engineering, with receipts.

An MCP server and CLI built on the FIBER decision-context compiler: a typed decision query is compiled into the smallest decision-sufficient evidence region, delivered with a Context Certificate stating exactly what was omitted.

Implementation of the AURORA BioPRISM / OncoWorld / FIBER blueprint (v0.6, 935 registered spec modules). A Rust workspace whose central idea is that context assembly is a compiler pass: instead of retrieval plus summarisation plus vibes-based compaction, a typed decision query is compiled into the smallest decision-sufficient evidence region, delivered as a Decision Section, and accompanied by a Context Certificate that states exactly what was omitted and whether the omission could have changed the decision.

Compile the smallest decision-sufficient evidence region. Never traverse the whole knowledge structure by default.

What the measurements actually say

The reference world ships 761 facts, 750 of them exploratory distractors that all consume the same protected cohort_id hub. FIBER compiles the query down to 11 facts (1.45% of the world) and the deterministic oracle still returns the correct verdict with all four leakage witnesses.

It is not alone in doing so. Under equal tuning, a 5-hop incidence walk and a BM25 retriever at k=11 select exactly the same eleven facts. The distribution's own compare_baselines.py measures the graph baseline only at depth 7 and unbounded — the two settings where it returns everything — and reports a 69× advantage that vanishes under equal tuning. That is a strawman comparison, and correcting it is what 43.38 and 43.41 require.

So the reference world cannot tell these methods apart. crates/worldgen makes the structure a parameter and builds one that can — distractors attached near the target instead of at a hub leaf, decisive facts behind a relay chain, and distractor tags camouflaged to tokenise into the protected vocabulary:

Three distinct failure modes appear. The graph walk has no usable depth: 5–10 pull in 98% of the world and still miss every decisive witness; 11 is the first sound setting and by then it has taken everything. BM25 reaches the right verdict from a 91% protected closure — right by luck, having dropped a protected fact that happened not to matter, and raising k to 50 never recovers it. FIBER is the only admissible strategy: right verdict and full closure, at 11 facts.

That last failure is why the harness ranks on admissibility rather than verdict alone — ranking on verdict would have crowned the strategy that violated the mandatory closure and got away with it.

This does not show FIBER wins generally: the discriminating world was built to expose these modes, just as the reference world was built to expose hub expansion. The full structural family sweep has now been run — 36 cells over attachment x relay depth x tag style x distractor count — and the two formerly missing baselines are in the panel. The sweep's headline is a negative result for FIBER: a plain backward walk over the directed factor edges, closure first, is admissible in all 36 cells at exactly FIBER's fact count, so on this family admissibility and cost cannot distinguish the compiler from that walk; the fixed-basis embedding retriever, by contrast, fails every camouflaged cell at the tight budget. Full analysis: docs/FINDINGS.md. How much of the blueprint the workspace actually covers, and which sections have nothing standing in for them: docs/COVERAGE.md. The crate layout and the blueprint path: docs/ARCHITECTURE.md.

Autonomous agent process boundary

The Python SDK includes a secret-safe operator entry point for the autonomous brain:

cd python
python -m prism_sdk catalogue
python -m prism_sdk evidence-plan --domain science
python -m prism_sdk provider-status --provider openai

For keyless local development, the same boundary supports an explicit credentialless fixture: python -m prism_sdk provider-status --provider local and run --provider local --model local-model use the runtime's bounded in-memory transport; no key or network provider is contacted.

Use run with a caller-owned MCP server when you are ready to invoke a provider. Keys are accepted only through a hidden prompt or an explicitly named environment variable; they are never command line arguments, MCP arguments, plans, or persisted state. See the autonomous brain guide for model discovery, durable inventory refresh, model-selection, approval, and credential-lifecycle details.

For post-run operations, both SDKs expose digest-bound, metadata-only trace analytics through analyze_autonomous_run_trace() / analyzeAutonomousRunTrace() and the corresponding agent facade methods. The report separates measured values from unmeasured domains, aggregates provider/model failure and latency observations, and emits conservative threshold alerts; it does not infer cost, task correctness, provider health, or domain truth. Longitudinal deployments can retain validated reports through the bounded AutonomousRunAnalyticsLedger with digest-checked restore and optional CAS persistence. See the analytics section.

Status

83 crates, 538,938 lines, clippy -D warnings enforced in CI. Byte-level parity with the CPython reference runtime is enforced by test and holds across three implementations: CPython, the Rust eager path, and the Rust indexed store.

The table below is generated. It used to be hand-maintained and drifted to claiming twenty-three crates and 820 tests — the same hand-copy drift crates/devx's exit-code audit exists to catch, sitting in the README of the repository that wrote the audit. Regenerate it, and the test count, with:

tools/status.sh --tests

The Blueprint column is derived rather than declared: it lists the sections whose module ids a crate actually cites in its own source, using the token rule tools/coverage.sh runs. A crate that stops citing a section drops it here without anyone remembering to edit a row.

How much of the blueprint is covered, and what the remainder is: docs/COVERAGE.md and docs/BACKLOG.md. Every uncovered module carries a typed verdict in crates/residue explaining why nothing implements it.

Read the full README on GitHub →