Pillar F: LYRA Bridge Integration Prompts

Seven prompts for exercising the v1.5.0 stable LYRA bridge — fcc.knowledge.lyra_bridge — across graph-of-thought queries, harmonic SKOS expansion, sentinel-guarded fallback, and eventual live-backend promotion.

LYRA is Orpheus's lyre: an instrument of harmony. The bridge gives FCC GraphRAG a vocabulary-expansion co-processor via broad-match chains. Currently the mock bridge walks a 7-node / 6-edge in-memory seed graph; the live lyra package is pending.

Context for Claude

get_lyra_bridge() returns a stable interface: expand(term, hops), broad_match(term), harmony(a, b). KGSE owns the eventual live-backend promotion; DAR preserves the decision trail.

Prompt 1: Harmonic expansion

LYRA, expand the term "observability" by 2 hops. Walk:
- Starting term
- Hop-1 matches
- Hop-2 matches
- Stopping rule (depth / breadth)

Note whether the mock or live backend served the query.

Expected outcome: expansion tree.

Prompt 2: Integrate with Pillar A

Chain Pillar A (GraphRAG) → Pillar F (LYRA). When GraphRAG returns
<3 matches, trigger LYRA expansion. Show:
- Trigger condition
- Expansion output
- Re-ranked combined result
- Provenance annotation on each result (original vs expanded)

Expected outcome: chained retrieval result.

Prompt 3: Sentinel-guarded invocation

When `lyra` is not importable, the mock bridge serves. Show:
- Detection
- Fallback
- Observability
- User-facing message (if any)
- Test that both code paths behave identically at the interface

Expected outcome: identical-interface verification.

Prompt 4: Live-backend promotion readiness

Produce the live-backend promotion readiness checklist:
- Repo creation (owner-approval-gated)
- API parity with mock
- Test coverage delta
- Observability contract
- KGSE authorship sign-off
- DAR decision-trail dossier

Expected outcome: promotion readiness checklist.

Prompt 5: Query evaluation

Evaluate LYRA expansion quality on a 10-query benchmark. Report:
- Precision@10 before expansion
- Precision@10 after expansion
- Any query where expansion degraded quality
- Recommended default hop count

Expected outcome: quality evaluation.

Prompt 6: Harmonic similarity

LYRA's `harmony(a, b)` returns a similarity score in [0, 1] rooted
in shared broad-match ancestors. For 5 persona pairs, compute
harmony and explain the score:
- (BC, RC) — peer authors
- (BC, OTO) — author vs operator
- (SMC, SFBI) — both jv_governance
- (DAR, PCT) — both governance, different focus
- (Ophiuchus, Serpens) — paired constellations

Expected outcome: 5-pair harmony narrative.

Prompt 7: KG serialization with LYRA tags

Serialize a KG slice that includes LYRA expansion annotations
(broad-match hops). For each target serializer (OWL, RDF, SKOS,
JSON-LD):
- How is the annotation preserved?
- Is any information lost?
- Recommended default

Expected outcome: serialization with LYRA annotation integrity.

Related

• Tutorial: docs/tutorials/advanced-capabilities/lyra-knowledge-graph-integration.md (Pool B)
• Personas: KGSE, DAR, VDS
• Decoder: docs/ecosystem/codename-decoder.md (LYRA entry)
• Related prompts: knowledge-engineers-lyra-graph-queries.md, pillar-graphrag-prompts.md