feat: intent-routing example + approve_federation_agent command

[toadkicker]

Summary

• examples/intent-routing: New self-contained example demonstrating Papillon's real 2-level intent pipeline — URL fast-path (detect_intent) → BM25 semantic index (IntentIndex::classify) → honest DuckDuckGo fallback. Comments explicitly explain that BM25 is a term-frequency scorer and the prompts are cherry-picked catalog matches. The resolved action type drives a full 6-phase PAP handshake (mandate → delegation → session → co-signed receipt).
• DynamicAgentSource::Federation: New enum variant marking agents approved from federation peer registries. No local keypair (operator_key_seed = None) — the operator's DID is the signing authority. Fixes the non-exhaustive match in papillon-shared/src/db/native.rs.
• approve_federation_agent Tauri command: validates advertisement signature via verify_key_from_did + ad.verify(), converts to a DynamicAgentDef with source = Federation, persists to SQLite, and registers in the live local registry. Federation agents are immediately visible to IntentIndex (BM25) on the next IntentIndex::new() call — closing the gap where AgentAdvertisement objects from remote registries were invisible to the BM25 routing layer.

Test Plan

• cargo test -p pap-intent-routing — 6 tests pass (BM25 routing, URL fast-path, fallback, full handshake integration)
• cargo test -p pap-agents — dynamic_agent_source_variants_round_trip + federation_source_serializes_as_federation pass
• cargo test -p papillon --lib commands::agents — 7 tests pass including approved_federation_def_is_visible_to_bm25
• cargo test -p papillon -p papillon-shared -p pap-agents -p pap-intent-routing — 353 tests, 0 failures
• cargo check -p papillon — clean compile after rebase on main

🤖 Generated with Claude Code

[greptile-apps]

toadkicker has reached the 50-review limit for trial accounts. To continue receiving code reviews, upgrade your plan.

[github-actions]

Benchmark Regression Report

PAP Protocol Benchmark Regression Check
========================================
Baseline: .bench-baseline/baseline.json
Threshold: 55%

  ed25519_keypair_generation                19.7 µs  (baseline: 17.7 µs, +11.6%)  [ok]
  did_key_derivation                         1.5 µs  (baseline: 1.3 µs, +17.6%)  [ok]
  mandate_create_sign                       23.9 µs  (baseline: 20.8 µs, +14.8%)  [ok]
  mandate_chain_verify_depth3              159.4 µs  (baseline: 120.6 µs, +32.1%)  [ok]
  sd_jwt_issue_5claims                      27.7 µs  (baseline: 24.0 µs, +15.5%)  [ok]
  sd_jwt_verify_disclose_3of5               54.2 µs  (baseline: 41.4 µs, +30.7%)  [ok]
  session_open_full_lifecycle              112.1 µs  (baseline: 93.0 µs, +20.5%)  [ok]
  receipt_create_cosign                     48.5 µs  (baseline: 41.6 µs, +16.5%)  [ok]
  federation_announce_local                 62.1 µs  (baseline: 51.5 µs, +20.4%)  [ok]


P99 Tail-Latency Check
----------------------
Results: target/p99_results.json
Threshold: 50%

  session_open_full_lifecycle(p99)         120.2 µs  (baseline: 500.0 µs, -75.9%)  [ok]
  mandate_chain_verify_depth3(p99)         170.6 µs  (baseline: 480.0 µs, -64.4%)  [ok]
  receipt_create_cosign(p99)                56.2 µs  (baseline: 210.0 µs, -73.2%)  [ok]

All benchmarks within 55% of baseline.

_{Threshold: 10% regression vs baseline from main}

[greptile-apps]

toadkicker has reached the 50-review limit for trial accounts. To continue receiving code reviews, upgrade your plan.

[greptile-apps]

toadkicker has reached the 50-review limit for trial accounts. To continue receiving code reviews, upgrade your plan.