Feature Matrix
This matrix is the implementation contract for the current codebase. "Introduced" tracks the release where a public capability first became a documented surface.
| Area | Status | Introduced | Tags | Evidence |
|---|---|---|---|---|
| Embedded mode | Implemented | v0.1 | runtime | client.ModeEmbedded, memory stores, Badger persistence |
| Standard mode | Implemented | v0.1 | runtime | Postgres graph/KV/event/vector stores and migrations |
| Remote mode | Implemented | v0.1 | runtime | JSON-over-gRPC remote stores |
| Scaled mode | Unavailable | v0.2 | runtime, scale | Reserved configuration; current build rejects it explicitly. Use standard mode. |
| Score breakdown | Implemented | v0.3 | inspectability, non-breaking | SDK, REST, gRPC, GraphQL expose weighted score contributions |
| Write deduplication | Implemented, opt-in | v0.3 | cost, non-breaking | Content fingerprints skip repeat embedding and touch existing nodes when enabled per request or server |
| GraphQL search | Implemented | v0.3 | inspectability, product-surface | /graphql search, filters, edge and source resolvers |
| Source credibility | Implemented | v0.1 | epistemics | Beta credibility model, labels, admission gate |
| Conflict detection | Implemented | v0.1 | epistemics | Write-time semantic contradiction tracking |
| Narrative retrieval | Implemented | v0.3 | epistemics, inspectability | Go SDK, REST, and GraphQL expose structured narrative reports |
| Knowledge gaps | Implemented | v0.3 | epistemics, active-learning | Go SDK, REST, and GraphQL expose gap reports |
| Knowledge acquisition planner | Implemented | v0.9 | active-learning, epistemics, operations | Go SDK, REST, and GraphQL expose prioritized acquisition tasks from gaps and weak claims |
| Acquisition execution connectors | Implemented | v0.105 | active-learning, operations, source-constraints | Go SDK, REST, TypeScript, and Python expose dry-run-first search/crawler connector workflows with source allow-lists |
| Acquisition provider connectors | Implemented | v0.106 | active-learning, integrations, providers | contextdb connectors serve exposes OpenAI, xAI, and Anthropic adapter endpoints for acquisition execution |
| Acquisition connector retry receipts | Implemented | v0.107 | active-learning, audit, reliability | Executed connector attempts record receipts, idempotency keys, retry classification, retry candidates, and backoff recommendations |
| Feedback APIs | Implemented | v0.3 | feedback-loop, non-breaking | Go SDK, REST, gRPC, and GraphQL expose validate/refute/useful/stale |
| Explain-rank | Implemented | v0.8 | inspectability, ranking, non-breaking | Go SDK, REST, and GraphQL compare two nodes with score component deltas |
| Explain-rank graph evidence | Implemented | v0.11 | inspectability, ranking, graph | Explain-rank responses include support-chain evidence and compound confidence |
| Feedback event log | Implemented | v0.5 | feedback-loop, audit, non-breaking | Go SDK, REST, and GraphQL expose durable feedback events |
| Source trust timeline | Implemented | v0.6 | audit, epistemics, feedback-loop | Go SDK, REST, and GraphQL expose credibility points from feedback events |
| Claim review queue | Implemented | v0.7 | feedback-loop, epistemics, operations | Go SDK, REST, and GraphQL expose ranked review tasks for refuted, stale, low-confidence, and contradictory claims |
| Central review worker | Implemented, opt-in | v0.123.0 | review, operations, audit | Server-owned dry-run-first rules and trusted webhook cycles persist decisions and manual-assignment lanes |
| Optional token registry | Implemented, opt-in | v0.123.0 | security, deployment | Exact configured bearer tokens protect REST, gRPC, observability, and admin surfaces; /health stays public |
| Recoverable write intents | Implemented | v0.123.0 | durability, audit | Pending-write inspection and recovery surfaces persist staged write intent; conflicts require manual recovery |
| Review workflow persistence | Implemented | v0.12 | feedback-loop, operations, audit | Go SDK, REST, and GraphQL expose append-only review decisions for assignment, snooze, resolution, and notes |
| Source trust anomaly alerts | Implemented | v0.13 | feedback-loop, epistemics, operations | Review queues emit source-trust anomaly tasks for credibility drops, low trust, and repeated refutations |
| Review queue filters | Implemented | v0.15 | feedback-loop, operations, inspectability | Go SDK, REST, and GraphQL filter review queues by task type, source, workflow status, and owner |
| Version and feature introspection | Implemented | v0.4 | introspection, non-breaking | REST /v1/version, /v1/features, /v1/migrations; GraphQL version, features, migrations |
contextdb doctor | Implemented, non-mutating checks | v0.4 | operations, introspection | CLI checks live REST ping, version, features, and migration metadata |
contextdb doctor --sample-write | Implemented, opt-in mutating probe | v0.4.1 | operations, durability | CLI writes a deduplicated probe node and verifies vector retrieval sees it |
contextdb doctor --backup-marker | Implemented, opt-in readiness check | v0.10 | operations, backup, durability | CLI verifies a backup marker file exists and is newer than --max-backup-age |
contextdb doctor --store-consistency | Implemented, opt-in local check | v0.49 | operations, durability | CLI samples valid graph nodes, checks fingerprint lookup, and reports vector rebuild candidates |
contextdb doctor --kv-key | Implemented, opt-in local check | v0.54 | operations, durability | CLI samples expected KV hot keys and reports missing cache refresh candidates |
contextdb doctor --kv-derived-key | Implemented, opt-in local check | v0.71 | operations, durability | CLI validates derived KV metadata and warns when generated_at is older than --max-kv-derived-age |
kv_derived_freshness repair hint | Implemented | v0.75 | operations, durability | Doctor failures include a dry-run repair kv-cache --derive recent-nodes command hint |
contextdb doctor --published-backup-url | Implemented, opt-in readiness check | v0.57 | operations, backup, durability | CLI reuses published backup catalog freshness checks inside the combined doctor report |
contextdb doctor --published-backup-index | Implemented, opt-in readiness check | v0.62 | operations, backup, durability | CLI reuses published backup catalog drift checks inside the combined doctor report |
contextdb doctor --published-backup-receipt | Implemented, opt-in readiness check | v0.82 | operations, backup, audit | CLI verifies published backup repair receipts against the local lifecycle index inside the combined doctor report |
| Doctor backup receipt runbook lane | Implemented | v0.95 | operations, backup, audit | Published backup repair docs now include a freshness, drift, receipt, doctor, and final closeout lane |
| Doctor backup receipt closure artifact bundle | Implemented | v0.98 | operations, backup, audit | Published backup repair docs define stable artifact filenames for dry-run, execute, receipt, receipt-check, and final doctor reports |
| Published backup closure bundle manifest | Implemented | v0.101 | operations, backup, audit | snapshot lifecycle index publish closure-bundle --dir --out writes a manifest with expected artifact paths, byte counts, and SHA-256 hashes |
| Published backup closure bundle verify | Implemented | v0.108 | operations, backup, audit | snapshot lifecycle index publish closure-bundle verify --manifest re-checks saved bundle byte counts and SHA-256 hashes |
published_backup_drift repair hint | Implemented | v0.66 | operations, backup, durability | Drift reports include a dry-run publish command hint when local catalog metadata should replace published metadata |
| Published backup repair guard | Implemented | v0.70 | operations, backup, durability | Deployment docs provide safety checks before executing published backup catalog replacement |
| Published backup repair receipt | Implemented | v0.74 | operations, backup, audit | snapshot lifecycle index publish --execute --receipt-out writes durable evidence for catalog replacement |
| Published backup repair receipt verify | Implemented | v0.78 | operations, backup, audit | snapshot lifecycle index publish receipt verify --receipt --in compares repair receipts against local lifecycle index payloads |
contextdb repair vector-index | Implemented, dry-run first | v0.50 | operations, durability | CLI reports vector rebuild candidates and reindexes reviewed graph-node vectors only with --execute |
contextdb repair kv-cache | Implemented, dry-run first | v0.59 | operations, durability | CLI plans reviewed KV hot-key refreshes and writes explicit values only with --execute |
contextdb repair kv-cache --derive recent-nodes | Implemented, dry-run first | v0.63 | operations, durability | CLI derives reviewed recent-node session context values from graph data before optional KV writes |
| KV derivation recipes | Implemented | v0.67 | operations, durability, deployment | Deployment docs provide naming, review, and promotion recipes for derived recent-node KV cache values |
| KV derived repair execution recipe | Implemented | v0.79 | operations, durability, deployment | Deployment docs provide a guarded doctor-hint to dry-run to execute checklist for stale derived KV refreshes |
| KV derived refresh receipt | Implemented | v0.83 | operations, durability, audit | repair kv-cache --derive recent-nodes --execute --receipt-out writes value hash and doctor confirmation evidence |
| KV derived refresh receipt verifier | Implemented | v0.96 | operations, durability, audit | repair kv-cache receipt verify --receipt --value-file validates receipt structure, embedded report evidence, doctor command, and optional reviewed value hash |
| KV derived refresh receipt doctor lane | Implemented | v0.99 | operations, durability, audit | contextdb doctor --kv-refresh-receipt --kv-refresh-value-file validates derived KV refresh receipts inside the combined health report |
| KV refresh receipt fixture bundle | Implemented | v0.102 | operations, durability, audit | Public docs fixtures include a valid derived KV refresh receipt/value pair and a mismatched value for CI verification |
| Public fixture catalog | Implemented | v0.108 | operations, durability, docs | /fixtures/index.json catalogs public KV receipt fixtures and their expected verifier outcome |
| Snapshot backup/restore | Implemented | v0.17 | operations, backup, durability | Go client and contextdb snapshot export/import provide NDJSON backup, seeded export filters, namespace override, and import dry-run validation |
| Snapshot restore reports | Implemented | v0.18 | operations, backup, inspectability | Go client report helpers and contextdb snapshot import --report summarize lines, records, vectors, and namespace overrides |
| Snapshot backup marker | Implemented | v0.19 | operations, backup, durability | contextdb snapshot export --backup-marker writes a doctor-compatible marker only after export succeeds |
| Snapshot diff preview | Implemented | v0.20 | operations, backup, inspectability | Snapshot restore reports classify nodes as new, changed, or unchanged during dry-run and import |
| Backup runbook | Implemented | v0.21 | operations, backup, deployment | Deployment docs provide scheduled export, restore preview, marker, doctor, launchd, systemd, and Norn check flow |
| Backup artifact manifest | Implemented | v0.22 | operations, backup, audit | contextdb snapshot export --manifest writes a checksummed JSON sidecar with namespace, version, bytes, marker path, and record counts |
| Backup manifest verify | Implemented | v0.23 | operations, backup, audit | contextdb snapshot verify --manifest --in validates checksum, size, and record counts before restore |
| Restore rehearsal | Implemented | v0.24 | operations, backup, audit | contextdb snapshot rehearse --manifest --in --namespace verifies the artifact and returns dry-run restore counts in one preflight report |
| Restore promotion checklist | Implemented | v0.25 | operations, backup, audit | Rehearsal reports include timestamp, target namespace, and a shell-quoted recommended import command |
| Restore promotion receipt | Implemented | v0.26 | operations, backup, audit | contextdb snapshot import --promotion-report --promotion-note writes a JSON receipt with import counts after successful promotion |
| Promotion receipt verify | Implemented | v0.27 | operations, backup, audit | contextdb snapshot receipt verify --promotion-report --manifest compares promotion receipts with artifact manifests |
| Backup lifecycle bundle | Implemented | v0.28 | operations, backup, audit | Backup runbook provides a guarded script for export, verify, rehearse, doctor, optional promotion, receipt verification, and lifecycle summary |
| Lifecycle summary verify | Implemented | v0.29 | operations, backup, audit | contextdb snapshot lifecycle verify --summary --report checks a lifecycle summary and its referenced backup, manifest, rehearsal, promotion, and receipt-check artifacts |
| Lifecycle retention report | Implemented | v0.30 | operations, backup, audit | contextdb snapshot lifecycle retention --dir --keep --report groups lifecycle bundles and marks newest artifacts to keep versus older pruneable bundles without deleting files |
| Lifecycle delete plan | Implemented | v0.31 | operations, backup, audit | contextdb snapshot lifecycle retention --emit-delete-script prints reviewed rm -- ... commands for pruneable artifacts without deleting files |
| Lifecycle manifest index | Implemented | v0.32 | operations, backup, audit | contextdb snapshot lifecycle index --dir --out --report writes a compact JSON catalog with bundle retention decisions, artifact sizes, and hashes |
| Lifecycle index verify | Implemented | v0.33 | operations, backup, audit | contextdb snapshot lifecycle index verify --in --report re-checks indexed artifact existence, byte sizes, and SHA-256 hashes |
| Lifecycle index diff | Implemented | v0.34 | operations, backup, audit | contextdb snapshot lifecycle index diff --old --new --report compares backup catalogs for bundle membership, retention decision, and artifact hash changes |
| Lifecycle index publish | Implemented | v0.36 | operations, backup, audit | contextdb snapshot lifecycle index publish --in --dry-run --report validates backup catalog metadata for ops publication without uploading backup contents |
| Lifecycle index publish drift | Implemented | v0.47 | operations, backup, audit | contextdb snapshot lifecycle index publish drift --in --published-url --report compares local and published backup catalog metadata |
| Lifecycle index publish freshness | Implemented | v0.52 | operations, backup, audit | contextdb snapshot lifecycle index publish freshness --published-url --max-age --report checks published catalog age |
| Release health page | Implemented | v0.11.2 | operations, release, durability | Docs page records unit, docs-build, ranking, durability, API contract, and race/soak release gates |
| CI-backed release health | Implemented | v0.108 | operations, release, ci | CI publishes JSON and Markdown release-health rows from test, build, durability, Postgres, and Docker job results |
| Hard durability CI lanes | Implemented | v0.108 | operations, durability, ci | CI runs race-sensitive packages, a concurrent soak test, and Docker-backed Postgres integration smoke coverage |
| Durability and ranking tests | Implemented | v0.4 | durability, ranking | Badger restart test, ranking golden fixtures, representative corpus ranking coverage, gRPC contract test, REST failure-path coverage |
| Ranking eval snapshots | Implemented | v0.48 | ranking, release, inspectability | contextdb eval ranking --out --report emits top-k, MRR, expected rank, and score breakdowns for the representative corpus |
| Ranking eval Markdown recap | Implemented | v0.55 | ranking, release, inspectability | contextdb eval ranking --markdown and --markdown-out emit pass/fail totals, MRR, failures, top results, and score breakdowns |
| Ranking eval historical diff | Implemented | v0.56 | ranking, release, inspectability | contextdb eval ranking --compare previous.json --diff-markdown emits MRR, pass, rank, and top-score movements between snapshots |
| Ranking eval baseline policy | Implemented | v0.61 | ranking, release, inspectability | contextdb eval ranking --baseline-dir writes versioned baseline artifacts and --compare-baseline-dir resolves the latest previous baseline |
| Ranking eval baseline retention | Implemented | v0.64 | ranking, release, inspectability | contextdb eval ranking --baseline-retention-dir reports current, retained, and pruneable baseline artifacts |
| Ranking eval baseline delete script | Implemented | v0.68 | ranking, release, inspectability | contextdb eval ranking --baseline-retention-dir --emit-delete-script emits a reviewed shell deletion plan for existing pruneable baseline artifacts without deleting files |
| Ranking baseline retention cookbook | Implemented | v0.72 | ranking, release, inspectability | Deployment docs provide keep-count, CI artifact, review, and delete-script recipes for ranking baseline history |
| Ranking baseline artifact manifest | Implemented | v0.76 | ranking, release, audit | contextdb eval ranking --baseline-manifest-out writes artifact inventory JSON with bytes and SHA-256 hashes |
| Ranking baseline artifact manifest verify | Implemented | v0.80 | ranking, release, audit | contextdb eval ranking baseline manifest verify --manifest verifies inventory paths, byte sizes, and SHA-256 hashes |
| Ranking baseline manifest verification Markdown | Implemented | v0.84 | ranking, release, audit | contextdb eval ranking baseline manifest verify --markdown-out writes compact verification recaps |
| Ranking manifest failure annotations | Implemented | v0.85 | ranking, release, ci | contextdb eval ranking baseline manifest verify --annotations and --annotations-out emit CI annotation lines for failed artifacts |
| Ranking manifest annotation workflow docs | Implemented | v0.86 | ranking, release, ci | Deployment docs include a GitHub Actions recipe for JSON, Markdown, and annotation artifacts |
| Ranking baseline verification bundle | Implemented | v0.87 | ranking, release, audit | contextdb eval ranking baseline manifest verify --bundle-dir DIR writes JSON, Markdown, and annotation files with stable names |
| Ranking bundle index verifier | Implemented | v0.90 | ranking, release, audit | contextdb eval ranking baseline manifest bundle verify --index PATH re-hashes bundle artifacts and checks JSON report status consistency |
| Mini/Norn deployment notes | Implemented | v0.3 | operations | Internal live deployment discovery and health-check docs |
| Norn registration helper | Implemented | v0.14 | operations, deployment | contextdb norn manifest and contextdb norn validate generate and check service entries |
| Norn live drift check | Implemented | v0.16 | operations, deployment | contextdb norn drift compares expected local service metadata with the live Norn manifest and reports field differences |
| Norn manifest publish | Implemented | v0.35 | operations, deployment | contextdb norn publish --dry-run --report validates a publish plan by default, with --execute --publish-url for explicit HTTP registration |
| Review escalation rules | Implemented | v0.37 | review, operations | Review queues can add escalation metadata for aged assigned or snoozed tasks and high-priority source anomaly items |
| Source quarantine workflow | Implemented | v0.108 | review, operations, trust | NamespaceHandle.SourceQuarantine dry-runs or executes label plans for repeatedly refuted, low-trust, or credibility-drifted sources |
| Review escalation digest | Implemented | v0.38 | review, operations | Review escalation digests group escalated tasks by owner, source, item type, and escalation level |
| Review escalation digest export | Implemented | v0.39 | review, operations | Review escalation digest snapshots can be recorded and listed for durable handoffs |
| Review handoff feed | Implemented | v0.40 | review, operations | Saved escalation digest snapshots can be polled by owner and escalation level |
| Review handoff webhook plans | Implemented | v0.41 | review, operations | Signed dry-run webhook delivery plans expose payload, headers, and retry metadata without sending requests |
| Review handoff webhook execution | Implemented | v0.42 | review, operations | Explicit webhook delivery sends synchronous POSTs with timeout, status, body, and error capture |
| Review handoff delivery receipts | Implemented | v0.43 | review, operations, audit | Executed handoff webhooks append durable receipts with target URL, status, payload hash, response hash, and errors |
| Review handoff retry candidates | Implemented | v0.44 | review, operations, audit | Failed handoff receipts are grouped by digest and target URL for retry review without sending retries |
| Review handoff retry execution | Implemented | v0.45 | review, operations, audit | Failed handoff candidates can be resent explicitly by digest event ID and target URL with a new receipt recorded |
| Review handoff retry backoff | Implemented | v0.46 | review, operations, audit | Failed handoff candidates include read-only retry pacing recommendations based on attempt history |
| Review handoff retry fatigue | Implemented | v0.51 | review, operations, audit | Retry recommendations are grouped by target endpoint with candidate, attempt, readiness, and status-family counts |
| Review handoff retry fatigue Markdown | Implemented | v0.53 | review, operations, handoff | REST retry-fatigue?format=markdown and Go helper render endpoint fatigue summaries for incident notes |
| Review handoff retry fatigue owner groups | Implemented | v0.58 | review, operations, handoff | Retry fatigue summaries include owner and escalation-level counts in JSON, GraphQL, and Markdown |
| Review handoff retry fatigue filters | Implemented | v0.60 | review, operations, handoff | REST, GraphQL, and Go SDK can filter retry fatigue by owner and escalation level |
| Review handoff retry fatigue cookbook | Implemented | v0.65 | review, operations, handoff | Deployment docs provide owner and escalation-lane recipes for retry fatigue handoffs |
| Review handoff retry fatigue presets | Implemented | v0.69 | review, operations, handoff | REST, GraphQL, and Go SDK support stable preset names for repeated owner and escalation lanes |
| Retry fatigue preset discovery docs | Implemented | v0.73 | review, operations, handoff | Deployment docs provide preset names, expanded filters, and intended handoff audiences for dashboards |
| Retry fatigue preset examples | Implemented | v0.77 | review, operations, handoff | Preset metadata includes copyable REST query and GraphQL argument snippets |
| Retry fatigue preset docs drift test | Implemented | v0.81 | review, operations, test | SDK tests verify cookbook preset rows match ReviewHandoffRetryFatiguePresets() metadata |
| Retry fatigue preset schema fixture | Implemented | v0.94 | review, operations, test | A shared JSON schema fixture guards SDK and REST preset payload fields, order, and required examples |
| Retry fatigue preset schema publication | Implemented | v0.97 | review, operations, docs | The preset JSON schema is published at /schemas/retry-fatigue-presets.schema.json and checked against the embedded fixture |
| Published schema catalog | Implemented | v0.100 | review, operations, docs | /schemas/index.json lists stable docs schemas with URLs, owners, feature names, status, and release provenance |
| Schema catalog drift annotations | Implemented | v0.108 | review, operations, ci | contextdb docs schema-catalog verify --annotations validates cataloged schema artifacts and emits CI annotation lines for drift |
| Admin/debug UI | Implemented | v0.88 | inspectability, operations | Observe port serves /admin/ with runtime stats and a belief debugger backed by /admin/api/belief |
| Admin debugger search | Implemented | v0.89 | inspectability, operations | /admin/api/search finds recent valid graph nodes by text, source, label, or ID and the UI can open them in the debugger |
| Admin metrics dashboard | Implemented | v0.91 | inspectability, operations, metrics | /admin/ surfaces health signals, ingest/retrieval rates, latency, and raw /admin/api/metrics JSON |
| Svelte admin shell | Implemented | v0.92 | inspectability, operations, ui | /admin/ is served from an embedded Svelte app bundle while preserving metrics, search, and belief debugger APIs |
| Debugger explain-rank compare | Implemented | v0.93 | inspectability, ranking, ui | /admin/api/explain-rank and the Svelte dashboard compare two nodes with rank summary and factor deltas |
| Ranking evaluation dashboard | Implemented | v0.103 | inspectability, ranking, ui | /admin/api/ranking-eval and /admin/ surface MRR, pass/fail, category health, query evidence, score components, and baseline deltas |
| Rich epistemics debugger visualization | Implemented | v0.104 | inspectability, epistemics, ui | /admin/api/belief includes an additive epistemics envelope and /admin/ renders source trust timelines, confidence history, contradiction paths, and graph/source context |
| Deferred acquisition review | Implemented, opt-in | v0.123.0 | review, acquisition | Candidates remain outside the graph until approved; repeat execution and admission are coordinated |
| Ranking result investigation | Implemented | v0.123.0 | ranking, ui | Evaluation IDs retain fixture context for direct debugger inspection |
| Browser ranking baselines | Implemented | v0.123.0 | ranking, ui | Validated bounded local history with selection, deletion, and storage-error handling |
| Current graph hydration | Implemented | v0.123.0 | retrieval, correctness | ANN candidates load current/as-of graph state before ranking |
Next Candidates
- Index pending-intent and idempotency-key lookups for large namespaces.
- Add candidate decisions, worker summaries, and manual recovery tasks to the reviewer UI.
- Extend Postgres failure injection and scheduled soak coverage.
- Filter debugger graph context by relation/source trust and add source quarantine affordances.
- Evaluate provider adapters on reviewed task corpora before enabling automatic evidence acquisition.
See the v0.123.0 recap for completed changes and the current release plan.