Phonton CLI v0.19.0

Phonton CLI v0.19.0

v0.19.0 ships an alpha slice of Phonton's next ADE architecture: typed swarm
planning, selectable code-index backends, and preview-only MCP capability
negotiation. It keeps the current product loop intact:

goal -> plan -> edit -> verify -> review -> remember

Added

• PlanGraph and SubtaskAssignment metadata in planner output, including
  role, expected touch scope, dependencies, conflict group, and verification
  intent.
• Broad-goal swarm activation with provider-backed planning when configured and
  visible deterministic fallback reasons when it is not.
• Conflict-aware dependency insertion in the orchestrator so overlapping touch
  scopes do not race while isolated subtasks still run concurrently.
• A CodeRetriever abstraction with default local-hnsw retrieval and an
  optional Qdrant HTTP backend for code and symbol retrieval.
• [index] config:

[index]
backend = "local-hnsw"

[index]
backend = "qdrant"
qdrant_url = "http://127.0.0.1:6333"
qdrant_collection = "phonton-code"

• phonton mcp capabilities <server-id> [--json] [--yes], which previews MCP
  initialize metadata, tool descriptors, and proposed permission rules without
  calling tools or writing config.

Changed

• phonton plan --json now includes swarm/index evidence fields.
• phonton doctor reports the selected index backend and checks Qdrant
  connectivity when Qdrant is configured.
• Flight Log telemetry includes MCP capability discovery evidence.
• Context and outcome evidence can carry plan graph and selected index backend
  data for review and future resume surfaces.

Guardrails

• Qdrant is an optional external code retrieval backend. Phonton does not start
  containers, manage collections, or migrate decision memory into Qdrant.
• SQLite keyword memory remains the authoritative local decision memory store.
• MCP capability negotiation is preview-only. It proposes sandbox permission
  rules for review but does not silently grant permissions or write config.
• v0.19.0 does not claim measured performance wins. Public benchmark claims
  still require reproducible fixtures, exact prompts, tool versions, provider
  token usage where available, raw logs, final diffs, verification logs, and
  handoff evidence.

Validation Checklist

cargo fmt --all -- --check
cargo test --locked --workspace
cargo clippy --locked --workspace --all-targets -- -D warnings
cargo build --release --locked -p phonton-cli
target/release/phonton.exe version
target/release/phonton.exe --help
npm run test:npm-wrapper
npm pack --dry-run