Live check dog-fooding: doc, code-gen and live-check

.github/workflows/ci.yml

@@ -67,6 +67,9 @@ jobs:
         run: cargo test --workspace --no-default-features
       - name: History check
         run: cargo xtask history
+      - name: Check generated files are up to date
+        if: matrix.os == 'ubuntu-latest'
+        run: cargo xtask check-generated
   msrv:
     name: Check MSRV
     runs-on: ubuntu-latest
@@ -279,7 +282,7 @@ jobs:
       - name: Install cargo-tarpaulin
         run: cargo install cargo-tarpaulin
       - name: Gather coverage
-        run: cargo tarpaulin --workspace --output-dir coverage --out lcov -e xtask -e weaver --exclude-files 'crates/weaver_forge/codegen_examples/expected_codegen/*'
+        run: cargo tarpaulin --workspace --output-dir coverage --out lcov --ignore-tests -e xtask -e weaver --exclude-files 'crates/weaver_forge/codegen_examples/expected_codegen/*' 'crates/weaver_live_check/tests/*'
       - name: Upload coverage reports to Codecov
         uses: codecov/codecov-action@671740ac38dd9b0130fbe1cec585b89eea48d3de # v5.5.2
         with:

CHANGELOG.md

@@ -9,6 +9,7 @@ All notable changes to this project will be documented in this file.
 - New feature ([#1100](https://github.com/open-telemetry/weaver/issues/1100)) - Set `--output=http` to have live-check send its report as the response to `/stop`. ([#1193](https://github.com/open-telemetry/weaver/pull/1193) by @jerbly)
 - Unified output handling across all registry subcommands. Builtin output formats (json, yaml, jsonl) are now available in `registry stats`, `registry diff`, and `registry resolve`. `registry stats` also supports `--templates` for custom text output templates. ([#1200](https://github.com/open-telemetry/weaver/pull/1200) by @jerbly)
 - New feature ([#1152](https://github.com/open-telemetry/weaver/issues/1152)) - Live-check with `--emit-otlp-logs` will now include the attributes from the resource in the emitted log record, this helps to identify the source of the finding in a multi-source environment. ([#1206](https://github.com/open-telemetry/weaver/pull/1206) by @jerbly)
+- Live-check OTLP log findings are now dog-fooded: the event schema, attributes, and enumerations are defined in a semconv model and code-generated using Weaver's own templates. See [`finding.md`](crates/weaver_live_check/docs/finding.md) for the generated reference documentation and [`dog-fooding.md`](crates/weaver_live_check/docs/dog-fooding.md) for the full dog-fooding guide. ([#1232](https://github.com/open-telemetry/weaver/pull/1232) by @jerbly)
 - New Experimental feature: `weaver registry infer` command that listens for OTLP telemetry and infers a semantic convention registry file from the received data. Supports spans, metrics, events, and resource attributes. Includes configurable gRPC address/port, admin server for graceful shutdown, and inactivity timeout. ([#1138](https://github.com/open-telemetry/weaver/pull/1138) by @ArthurSens)
 - Use `schema_url` in registry manifest, dependencies, and resolved schema instead of `registry_url`. Parse registry name and version
   from it. ([#1202](https://github.com/open-telemetry/weaver/pull/1202) by @lmolkova)

Cargo.toml

@@ -156,6 +156,7 @@ lto = "thin"
 [workspace.lints.rust]
 rust_2018_idioms = { level = "warn", priority = -1 }
 unused_lifetimes = "warn"
+unexpected_cfgs = { level = "warn", check-cfg = ['cfg(tarpaulin)', 'cfg(tarpaulin_include)'] }
 # unreachable_pub = "warn"
 semicolon_in_expressions_from_macros = "warn"
 missing_docs = "deny"

crates/weaver_checker/src/finding.rs

@@ -18,7 +18,7 @@ pub struct PolicyFinding {
     /// machine-readable string that categorizes the finding.
     pub id: String,
 
-    /// The context associated with the finding e.g. { "attribute_name": "foo.bar", "attribute_value": "bar" }
+    /// The context associated with the finding e.g. { "attribute_key": "foo.bar", "attribute_value": "bar" }
     /// The context should contain all dynamic parts of the message
     /// Context values may be used with custom templates and filters to customize reports.
     pub context: Value,

crates/weaver_live_check/Cargo.toml

@@ -26,10 +26,16 @@ opentelemetry_sdk.workspace = true
 opentelemetry-otlp.workspace = true
 opentelemetry-stdout.workspace = true
 tokio.workspace = true
+strum = { version = "0.27.2", features = ["derive"] }
 
 [dev-dependencies]
-tempfile = "=3.25.0"
+assert_cmd = "=2.1.2"
+portpicker = "=0.1.1"
 serde_yaml.workspace = true
+serial_test = "=3.2.0"
+tempfile = "=3.25.0"
+tokio = { workspace = true, features = ["macros", "rt-multi-thread"] }
+ureq.workspace = true
 
 [lints]
 workspace = true

crates/weaver_live_check/README.md

@@ -2,6 +2,12 @@
 
 Live check is a developer tool for assessing sample telemetry and providing findings for improvement.
 
+> **Dog-fooding**: Live Check uses Weaver's own semantic convention model and template engine to
+> define its `finding` schema, generate Rust types and constants, and generate reference
+> documentation — all from [`model/live_check.yaml`](model/live_check.yaml). An integration test
+> validates the emitted OTLP log records against the model, catching any drift between the
+> generated code and the schema. See [Dog-fooding](docs/dog-fooding.md) for details.
+
 A Semantic Convention `Registry` is loaded for comparison with samples. `Ingesters` transform various input formats and sources into intermediary representations to be assessed by `Advisors`. The `PolicyFinding` produced is transformed via jinja templates to the required output format for downstream consumption.
 
 ```mermaid
@@ -88,7 +94,7 @@ As mentioned, a list of `PolicyFinding` is returned in the report for each sampl
 - `signal_type`: _string_ - a type of the signal for which the finding is reported: `metric`, `span`, `log` or `resource`
 - `signal_name`: _string_ - a name of the signal for which the finding is reported: metric name, event name or span name
 - `context`: _any_ - a map that describes details about the finding in a structured way,
-  for example `{ "attribute_name": "foo.bar", "attribute_value": "bar" }`.
+  for example `{ "attribute_key": "foo.bar", "attribute_value": "bar" }`.
 - `message`: _string_ - verbose string describing the finding. It contains the same details as `context` but
   is formatted and human-readable.
 
@@ -100,7 +106,7 @@ As mentioned, a list of `PolicyFinding` is returned in the report for each sampl
         "level": "violation",
         "id": "missing_attribute",
         "message": "Attribute `hello` does not exist in the registry.",
-        "context": { "attribute_name": "hello" },
+        "context": { "attribute_key": "hello" },
         "signal_name": "http.client.request.duration",
         "signal_type": "metric"
       }
@@ -132,7 +138,7 @@ deny contains make_finding(id, level, context, message) if {
 	id := "contains_test"
 	level := "violation"
 	context := {
-		"attribute_name": input.sample.attribute.name
+		"attribute_key": input.sample.attribute.name
 	}
 	message := sprintf("Attribute name must not contain 'test', but was '%s'", [input.sample.attribute.name])
 }
@@ -254,6 +260,9 @@ weaver registry live-check --emit-otlp-logs --otlp-logs-stdout
 
 ### Log Record Structure
 
+> For the full auto-generated reference of finding attributes, enum values, and the event schema,
+> see [docs/finding.md](docs/finding.md).
+
 Each policy finding is emitted as an OTLP log record with the following structure:
 
 **Body**: The finding message (e.g., "Required attribute 'server.address' is not present.")
@@ -275,10 +284,10 @@ Each policy finding is emitted as an OTLP log record with the following structur
 - `weaver.finding.id`: Finding type identifier (e.g., "required_attribute_not_present")
 - `weaver.finding.level`: Finding level as string ("violation", "improvement", "information")
 - `weaver.finding.context.<key>`: Key-value pairs provided in the context. Each pair is recorded as a single attribute.
-- `weaver.finding.sample_type`: Sample type (e.g., "attribute", "span", "metric")
-- `weaver.finding.signal_type`: Signal type (e.g., "span", "event", "metric")
-- `weaver.finding.signal_name`: Signal name (e.g., event name or metric name)
-- `weaver.finding.resource_attribute.<key>`: Resource attributes from the original telemetry source that produced the finding
+- `weaver.finding.sample.type`: Sample type (e.g., "attribute", "span", "metric")
+- `weaver.finding.signal.type`: Signal type (e.g., "span", "metric", "log")
+- `weaver.finding.signal.name`: Signal name (e.g., event name or metric name)
+- `weaver.finding.resource.attribute.<key>`: Resource attributes from the original telemetry source that produced the finding
 
 ## Continuous Running Sessions
 

crates/weaver_live_check/data/policies/live_check_advice/test.rego

@@ -9,7 +9,7 @@ deny contains make_advice(advice_type, advice_level, advice_context, message) if
     advice_type := "contains_test"
     advice_level := "violation"
     advice_context := {
-        "attribute_name": input.sample.attribute.name
+        "attribute_key": input.sample.attribute.name
     }
     message := sprintf("Attribute name must not contain 'test', but was '%s'", [input.sample.attribute.name])
 }