pulseengine · avrabe · Jun 25, 2026 · Jun 25, 2026
diff --git a/README.md b/README.md
@@ -76,7 +76,7 @@ cargo install --path rivet-cli
 
 # Or grab a prebuilt release binary directly (substitute your platform /
 # the current version — assets are named rivet-vX.Y.Z-<target>)
-curl -L https://github.com/pulseengine/rivet/releases/download/v0.16.1/rivet-v0.16.1-x86_64-unknown-linux-gnu.tar.gz | tar xz
+curl -L https://github.com/pulseengine/rivet/releases/download/v0.19.0/rivet-v0.19.0-x86_64-unknown-linux-gnu.tar.gz | tar xz
 ```
 
 `npm`/`npx` is the zero-toolchain path — it fetches the same signed
@@ -87,7 +87,7 @@ mcp` without a Rust install.
 
 ```bash
 rivet init --preset dev          # 1. scaffold rivet.yaml + schemas/ + artifacts/
-rivet add requirement -t "DB write returns ack" --asil B   # 2. add a typed atom
+rivet add -t requirement --title "DB write returns ack"   # 2. add a typed atom
 rivet validate                   # 3. oracle: PASS (no diagnostics)
 rivet serve --port 3099          # 4. dashboard at http://localhost:3099
 rivet docs quickstart            # 5. 10-step embedded guide w/ oracles

diff --git a/docs/getting-started.md b/docs/getting-started.md
@@ -446,6 +446,66 @@ Orphan artifacts (no links): 3
   REQ-012
 ```
 
+### `rivet coverage`
+
+Report traceability coverage per rule — for each source type governed by a
+traceability rule, the share of artifacts that satisfy it.
+
+```bash
+rivet coverage
+rivet coverage --format json
+rivet coverage --fail-under 80          # non-zero exit if overall < 80%
+rivet coverage --tests                  # test-to-requirement coverage from source markers
+```
+
+For every source type governed by more than one rule, the report also prints a
+combined **V-closure** figure: the share of artifacts satisfying *every*
+applicable rule (e.g. requirements that are BOTH satisfied AND verified). This
+intersection is strictly stronger than the per-rule numbers, which hide the gap
+when different artifacts miss different rules. `--format json` carries it as a
+`closure` array (`closed` / `total` / `open_ids` / `percentage` / `rule_names`).
+External-boundary (supplier-delegated) artifacts count as closed, matching the
+3-state coverage convention.
+
+### `rivet verify`
+
+Advance a requirement from `implemented` to `verified` when it has verifying
+test evidence — an incoming `verifies` link OR a `// rivet: verifies <ID>`
+source marker (scanned from `src/` and `tests/`, or extra paths via `--scan`).
+
+```bash
+rivet verify REQ-042
+rivet verify REQ-042 --scan integration-tests/
+```
+
+This is an opt-in, auditable command: it refuses with an actionable message
+when there is no evidence, no-ops if the artifact is already `verified`, and
+rejects non-`implemented` states. The write reuses the schema-validated
+`modify --set-status` path.
+
+### `rivet query`
+
+Run an s-expression filter and print matches. Mirror of the MCP `rivet_query`
+tool and the `{{query:(...)}}` document embed — all three share one evaluator,
+so their results agree for the same filter. See
+[S-Expression Filtering](#s-expression-filtering) for the predicate reference.
+
+```bash
+rivet query '(= type "requirement")'
+rivet query '(and (= type "requirement") (has-tag "safety"))' --format json
+rivet query '(not (linked-from "verifies" _))' --format ids   # newline-separated IDs
+```
+
+### `rivet sql`
+
+Run SQL over the artifact store — reads and a constrained set of writes — with
+no server and no MCP required. See [SQL over the artifact
+store](#sql-over-the-artifact-store) for the table schema and examples.
+
+```bash
+rivet sql "SELECT id, status FROM artifacts WHERE type='requirement'"
+```
+
 ### `rivet matrix`
 
 Generate a traceability matrix showing coverage between two artifact types.
@@ -494,9 +554,16 @@ Export all project artifacts to a specified format.
 ```bash
 rivet export --format reqif --output artifacts.reqif
 rivet export --format generic-yaml                       # stdout
+rivet export --format html --output ./dist               # static dashboard
+rivet export --format zola --output ./site               # Zola content tree
+rivet export --format gherkin --output ./features        # .feature files
 ```
 
-Supported formats: `reqif` (ReqIF 1.2 XML), `generic-yaml`.
+Supported formats: `reqif` (ReqIF 1.2 XML), `generic-yaml`, `html` (static
+dashboard; `--single-page`, `--theme dark|light`, `--offline`), `zola` (a Zola
+content tree), and `gherkin` (`.feature` files). The `html`, `zola`, and
+`snapshot`/`embed` exporters share the dashboard renderer and are gated behind
+the `serve` cargo feature (present in the default build).
 
 ### `rivet serve`
 
@@ -863,6 +930,66 @@ on links, use the purpose-built predicates (`linked-by`, `linked-from`,
 
 ---
 
+## SQL over the artifact store
+
+`rivet sql` projects the artifact store into an in-memory SQLite database and
+runs SQL against it — **no server and no MCP required**. The projection is
+rebuilt per invocation, so results are never stale. SQL gives you JOINs and
+aggregations the s-expression filter cannot express; reach for it when a
+question spans relationships between artifacts.
+
+Four virtual tables are projected:
+
+| Table        | Columns                                                                  |
+|--------------|--------------------------------------------------------------------------|
+| `artifacts`  | `id`, `type`, `title`, `description`, `status`, `fields_json`            |
+| `links`      | `source`, `link_type`, `target`, `external`                              |
+| `fields`     | `artifact_id`, `key`, `value` (EAV — one row per field, for JOINs)       |
+| `provenance` | `artifact_id`, `created_by`, `model`, `session_id`, `timestamp`, `reviewed_by` |
+
+Output is `table` (default), `json` (one object per row, keyed by column), or
+`csv`.
+
+### Reads — SELECT / WITH
+
+```bash
+# Count artifacts by type
+rivet sql "SELECT type, COUNT(*) AS n FROM artifacts GROUP BY type ORDER BY n DESC"
+
+# The V-closure gap: implemented requirements with no incoming `verifies` link
+rivet sql "SELECT id FROM artifacts
+           WHERE type='requirement' AND status='implemented'
+             AND id NOT IN (SELECT target FROM links WHERE link_type='verifies')"
+
+# Join artifacts to their links
+rivet sql "SELECT a.id, l.link_type, l.target
+           FROM artifacts a JOIN links l ON l.source = a.id
+           WHERE a.type='requirement'" --format json
+
+# Provenance rollup
+rivet sql "SELECT created_by, COUNT(*) FROM provenance GROUP BY created_by"
+```
+
+### Writes — UPDATE
+
+A constrained write slice routes through rivet's validated mutation path:
+`UPDATE artifacts SET {status|title|description}` only. Every change is
+schema-validated (status enums, reserved keys) *before* any file is touched, so
+an invalid value aborts the whole statement — never a silent partial write —
+and the safe YAML editor preserves sibling fields. `INSERT`, `DELETE`, and
+writes to `fields` or `links` are refused with a clear message (use the
+dedicated `add` / `link` / `modify` commands for those).
+
+```bash
+# Promote a requirement once it has verifying evidence
+rivet sql "UPDATE artifacts SET status='verified' WHERE id='REQ-042'"
+```
+
+A no-op write (the value already matches) reports `0 artifacts changed` and
+touches nothing.
+
+---
+
 ## Variant Management (Product Line Engineering)
 
 Manage product variants with feature models, constraint solving, and artifact scoping.
@@ -1155,15 +1282,16 @@ The importer:
 
 ## MCP Server (AI Agent Integration)
 
-Rivet exposes 15 tools via the Model Context Protocol:
+Rivet exposes 16 tools via the Model Context Protocol:
 
 ```bash
 rivet mcp  # stdio transport
 ```
 
-Tools: `rivet_validate`, `rivet_list`, `rivet_get`, `rivet_stats`, `rivet_coverage`,
-`rivet_schema`, `rivet_embed`, `rivet_snapshot_capture`, `rivet_add`, `rivet_query`,
-`rivet_modify`, `rivet_link`, `rivet_unlink`, `rivet_remove`, `rivet_reload`.
+Tools: `rivet_validate`, `rivet_list`, `rivet_get`, `rivet_bundle`, `rivet_stats`,
+`rivet_coverage`, `rivet_schema`, `rivet_embed`, `rivet_snapshot_capture`,
+`rivet_add`, `rivet_query`, `rivet_modify`, `rivet_link`, `rivet_unlink`,
+`rivet_remove`, `rivet_reload`.
 
 All mutations are audit-logged to `.rivet/mcp-audit.jsonl`.
 

diff --git a/docs/what-is-rivet.md b/docs/what-is-rivet.md
@@ -193,10 +193,10 @@ reviews*.
 - **Question.** Can my AI agent read the trace graph and propose
   mutations that rivet server-side-validates before I accept them?
 - **Artifacts.** All of them, through MCP.
-- **AI does.** Calls 15 MCP tools: `rivet_list`, `rivet_get`,
-  `rivet_query`, `rivet_stats`, `rivet_coverage`, `rivet_schema`,
-  `rivet_embed`, `rivet_add`, `rivet_modify`, `rivet_link`,
-  `rivet_unlink`, `rivet_remove`, `rivet_validate`,
+- **AI does.** Calls 16 MCP tools: `rivet_list`, `rivet_get`,
+  `rivet_bundle`, `rivet_query`, `rivet_stats`, `rivet_coverage`,
+  `rivet_schema`, `rivet_embed`, `rivet_add`, `rivet_modify`,
+  `rivet_link`, `rivet_unlink`, `rivet_remove`, `rivet_validate`,
   `rivet_snapshot_capture`, `rivet_reload`.
 - **Human reviews.** The MCP audit log (`rivet mcp` writes every
   mutation call as JSONL to the project directory) plus the git diff.

diff --git a/rivet-cli/src/docs.rs b/rivet-cli/src/docs.rs
@@ -1,3 +1,3 @@
 //! `rivet docs` — built-in searchable documentation.
 //!
 //! All documentation is embedded in the binary. Topics are searchable
@@ -535,16 +535,32 @@
 rivet list [-t TYPE]        List artifacts (filter by type/status)
 rivet get ID                Show a single artifact by ID (text/json/yaml)
 rivet stats                 Summary statistics and orphan detection
-rivet coverage              Traceability coverage report
+rivet coverage              Traceability coverage report (incl. combined V-closure)
+rivet verify ID             Advance implemented -> verified on test evidence
 rivet matrix --from X --to Y  Traceability matrix between types
 rivet diff                  Compare artifact versions
-rivet export -f FORMAT      Export to reqif, generic-yaml, or html
+rivet export -f FORMAT      Export to reqif, generic-yaml, html, zola, or gherkin
 rivet serve [-P PORT]       Start HTMX dashboard (default: 3000)
 rivet commits [--since N]   Commit-artifact traceability analysis
 rivet commit-msg-check F    Validate commit message trailers (hook)
 rivet impact --since REF    Change impact analysis (transitive)
 ```
 
+## Query Commands
+
+```
+rivet query SEXPR           Run an s-expression filter (mirrors rivet_query)
+rivet sql "SELECT ..."      SQL over the artifact store — no server/MCP needed
+rivet sql "UPDATE artifacts SET status=... WHERE ..."   Validated write slice
+```
+
+`rivet sql` projects the store into in-memory SQLite (tables `artifacts`,
+`links`, `fields`, `provenance`) so JOINs and aggregations the s-expression
+filter can't express become one query. Reads are `SELECT`/`WITH`; the write
+slice is `UPDATE artifacts SET {status|title|description}` only, validated
+before any file is touched. See `rivet docs cli` examples in
+`docs/getting-started.md`.
+
 ## Embed and Snapshot Commands
 
 ```
@@ -566,7 +582,7 @@
 
 Exposes rivet tools to AI agents via the Model Context Protocol.
 The server uses stdio transport and only binds to the local process.
-See `rivet docs mcp` for the wire format, the 15-tool catalog, and the
+See `rivet docs mcp` for the wire format, the 16-tool catalog, and the
 3-message handshake.
 
 ## Oracle Subcommands (`rivet check`)
@@ -589,9 +605,11 @@
 
 ```
 rivet schema list           List all artifact types
+rivet schema presets        List declarable built-in schema presets (no project needed)
 rivet schema show TYPE      Show type details with example YAML
 rivet schema links          List all link types with inverses
 rivet schema rules          List all traceability rules
+rivet schema sources        Show where each active schema resolves (on-disk vs embedded)
 rivet schema migrate TGT    Plan + apply preset migration with snapshot
                             (see rivet docs schema-migrate)
 ```
@@ -620,7 +638,14 @@
 ```
 
 Available presets: `dev`, `aspice`, `stpa`, `stpa-ai`, `cybersecurity`,
-`aadl`, `eu-ai-act`, `safety-case`.
+`aadl`, `eu-ai-act`, `safety-case`, `do-178c`, `en-50128`, `iec-61508`,
+`iec-62304`, `iso-pas-8800`, `sotif`. Run `rivet schema presets` for the live
+list with versions and type counts.
+
+Init also accepts `--vendor-schemas` to write the resolved built-in schemas
+(plus auto-discovered bridges) on-disk into `schemas/`, pinning a project's
+validation against rivet upgrades. The loader prefers on-disk schemas over the
+embedded copies, so a vendored set is immune to release-to-release rule drift.
 
 ## Mutation Commands
 
@@ -2009,7 +2034,7 @@
 
 ```
 $ rivet mcp --list-tools
-rivet MCP server — 15 registered tools
+rivet MCP server — 16 registered tools
 
   rivet_add
     Add a new artifact to the project via CST mutation. Call rivet_reload after.
@@ -2572,6 +2597,12 @@
 source. Run `rivet validate` to detect drift; run `rivet check sources
 --update` to refresh the stamps.
 
+It is a `common` base field, so **any** artifact type can carry it without
+tripping `unknown-field` — including verification artifacts (`sw-req`,
+`unit`, `sys-verification`), which is exactly where a tamper-evident citation
+on the requirement→test→evidence chain is wanted. The drift checker discovers
+the field by name, so it runs on those types too.
+
 Phase 1 (current) implements the `kind: file` backend only. Remote kinds
 (`url`, `github`, `oslc`, `reqif`, `polarion`) are recognised by the
 schema and round-trip cleanly, but their backends ship in Phase 2.