Skip to content

MODULE.md

Latest commit

 

History

History
67 lines (57 loc) · 4.82 KB

MODULE.md

File metadata and controls

67 lines (57 loc) · 4.82 KB
Raw

:storage

Responsibility

Owns KNet's desktop persistence implementation and schema.

Owns

  • Room database, schema-v25 entities, DAOs, and storage data sources.
  • The current 14-to-15, 15-to-16, 16-to-17, and 17-to-18 migrations; unsupported older development schemas may still be reset.
  • Durable metadata records and references to externally stored bodies.
  • Canonical session/connection/exchange/body/message/annotation/gap/deletion-outbox records and indexed single-session or global keyset queries.
  • SQLite-generated immutable exchange capture sequences, exact filtered-count queries, and sequence-keyed paging indexes that keep ordering stable as pages are loaded.
  • Single-exchange detail and bounded multi-exchange Flow queries for semantic annotations, allowing list presentation to react to post-capture inspection without reading message bodies.
  • Bounded exchange batches used by import and the 100,000-row indexed-query regression fixture.
  • Aggregate and oldest-first projections used by bounded global count/byte retention and startup lifecycle/body recovery.
  • Non-null opaque body storage keys and indexed ownership queries used for finalized-orphan reconciliation.
  • Durable corrupt/missing body states used by bounded integrity verification and fail-safe presentation.
  • Registered-device identities, non-plaintext trusted credential material, and one-shot pairing invitations.
  • Breakpoint rule rows containing transport filters plus a generic normalized protocol ID and opaque extension-owned criteria payload and deterministic priority; storage never decodes protocol semantics.
  • Complete API Studio authored-request rows, including JSON-encoded query/header/cookie/body-field state.
  • Opaque API Studio workspace documents and imported/reflected schema sources. Storage retains editor identity, semantic request kind, location, name ownership, badge, payload version, and bytes, but never decodes protobuf, gRPC, WebSocket, or another editor document.

Does not own

  • Canonical traffic models, use cases, UI, engine lifecycle, or unbounded in-memory body retention.

Dependency rule

Expose storage-oriented APIs to data adapters; do not depend on UI or engine modules.

Current state

Schema v15 adds registered identities, trusted-device credential digests/public keys, and pairing invitations to the existing canonical capture and API Studio schema. Schema v16 adds the global timestamp/ID index used by cross-session traffic history. Its DAO performs keyset paging plus host/path/method/status search and typed method/status/scheme/effective-protocol filtering without materializing a session. Both auto-migrations preserve supported existing records. Plain pairing secrets and issued credentials are never stored. Unsupported earlier development databases are deliberately reset by Room instead of imported. Schema v17 completes API Studio request persistence with query parameters, cookies, raw-body format, and structured form fields. Draft promotion uses one DAO transaction so the saved row and draft deletion cannot be observed partially. Schema v18 adds request-title ownership. Existing rows default to USER_DEFINED, preventing generated naming from changing any request title created before this capability. Schema v19 makes the generated capture sequence the exchange row key and keeps canonical ExchangeId under a unique index. KNet is still in development, so the existing destructive-development fallback recreates an unsupported v18 database instead of carrying a compatibility migration; the exported v19 schema is canonical. Schema v20 adds the API Studio request's HTTP-version preference. Existing v19 authored requests migrate to AUTO; future unknown stored tokens are also interpreted as AUTO by the data adapter. Schema v21 adds canonical traffic-origin attribution to exchanges. Existing captures migrate to proxy-client; request and response protocol columns continue to preserve the two observed protocol legs. Schema v22 adds independently encoded request and response trailer columns. Existing captures default to empty trailers; the already-canonical stream ID and two protocol-leg columns require no compatibility model. Schema v23 activates durable, sequence-keyed duplex protocol messages with direction, compression facts, bounded body ownership, terminal state, and error metadata. Schema v24 introduced the first opaque API Studio protocol document and schema-source records. Schema v25 replaces that prototype document row with the neutral api_studio_workspace_documents table, allowing incomplete drafts and saved-collection placement for any contributed editor while keeping schema sources independent. This development schema intentionally requires no compatibility model for the discarded v24 prototype rows. Breakpoint queries return priority then ID order so persistence and live rule evaluation share one stable order.