@tangle-network/traces 0.8.19 → 0.8.21

package/README.md

@@ -1,6 +1,6 @@
 # traces
 
-> Point `traces` at the session logs your coding agent already writes to disk — Claude Code, Codex, OpenCode, Gemini, and more — and get failure-mode + efficiency findings. **Zero instrumentation.** A CLI *and* an SDK.
+> Point `traces` at the session logs your coding agent already writes to disk: Claude Code, Codex, OpenCode, Gemini, and more. Get failure-mode and efficiency findings with zero instrumentation. A CLI *and* an SDK.
 
 ![traces analyzing a real Claude Code session](https://raw.githubusercontent.com/tangle-network/traces/main/docs/demo.gif)
 
@@ -8,7 +8,7 @@
 [![license](https://img.shields.io/npm/l/@tangle-network/traces.svg)](./LICENSE)
 [![node](https://img.shields.io/node/v/@tangle-network/traces.svg)](https://nodejs.org)
 
-It reads the transcripts your harness leaves on disk, reconstructs the run as spans, and reports where the agent got stuck, burned tokens, or stopped checking its own work — locally, with no API key and no cost for the deterministic pass.
+It reads the transcripts your harness leaves on disk, reconstructs the run as spans, and reports where the agent got stuck, burned tokens, or stopped checking its own work. The deterministic pass runs locally, with no API key and no cost.
 
 ## Contents
 
@@ -46,16 +46,16 @@ Requires Node ≥ 22.
 traces analyze --harness claude-code --last 1
 traces improve --harness claude-code --last 5 --dir .traces/improvement
 traces watch --all
-traces stream --all --no-spans
+traces stream --all --mode findings
 ```
 
-That's the command in the demo above. The **deterministic pass** — stuck loops, token growth, output decay, missing self-verification, tool monoculture — needs no API key and costs nothing.
+That's the command in the demo above. The **deterministic pass** checks stuck loops, token growth, output decay, missing self-verification, and tool monoculture. It needs no API key and costs nothing.
 
 Add `--llm` for the **agentic analysts** (failure-mode / knowledge-gap / knowledge-poisoning / improvement); they call OpenAI and respect `--budget <usd>`.
 
-Every run also writes a **canonical OpenInference JSONL artifact**, so you can run external engines like [HALO](https://github.com/context-labs/halo) over it directly with `--analyzer halo` — see [External engines](#external-engines-bring-your-own). Analysis is never locked to one engine.
+Every run also writes a **canonical OpenInference JSONL artifact**, so you can run external engines like [HALO](https://github.com/context-labs/halo) over it directly with `--analyzer halo`. See [External engines](#external-engines-bring-your-own). Analysis is never locked to one engine.
 
-`traces improve` is the reviewable action path. It writes typed artifacts — findings, recommendations, evidence rows, claims, report, and before/after replay metadata — so another agent, CI job, or hosted product can consume the result without scraping prose.
+`traces improve` is the reviewable action path. It writes typed artifacts: findings, recommendations, evidence rows, claims, report, and before/after replay metadata. Another agent, CI job, or hosted product can consume the result without scraping prose.
 
 ## What it finds
 
@@ -64,10 +64,10 @@ The deterministic pass (free, no key) surfaces:
 | Finding | Meaning |
 |---|---|
 | **Stuck loops** | the same tool called N× with identical args and no state change |
-| **Monotonic input growth** | full history re-sent every step — context never compressed |
+| **Monotonic input growth** | full history re-sent every step; context never compressed |
 | **Output-length decay** | planning/reasoning per step shrinking as context grows |
 | **No self-verification** | state-mutating actions never followed by an eval/inspect/check |
-| **Tool monoculture / retry / error rates** | the shape of how the agent actually spent its calls |
+| **Tool mix / retry / error rates** | repeated, retried, and failed tool calls |
 
 `--llm` adds agentic analysts that read the conversation and cluster higher-order failure and improvement signals.
 
@@ -88,7 +88,7 @@ The deterministic pass (free, no key) surfaces:
 | `github-copilot` (`copilot`) | `~/.copilot/session-state/<id>/events.jsonl` | fixture |
 | `forge` (`forgecode`) | `/dump` JSON exports | fixture |
 
-Every adapter captures the full conversation — the **user's prompt** and the **assistant's response** text, plus tool calls/results and token usage. (`github-copilot` is the one exception: its log format carries no user prompt.) Factory stores token totals in `.settings.json`, not per turn. Forge reads `/dump` JSON exports (live SQLite is a follow-up). ACP-only bridges may not persist a local transcript.
+Every adapter captures the full conversation: the **user's prompt** and the **assistant's response** text, plus tool calls/results and token usage. (`github-copilot` is the one exception: its log format carries no user prompt.) Factory stores token totals in `.settings.json`, not per turn. Forge reads `/dump` JSON exports (live SQLite is a follow-up). ACP-only bridges may not persist a local transcript.
 
 ## CLI reference
 
@@ -118,7 +118,7 @@ traces upload   --since 24h                        # upload last day to the Inte
 | `--last <n>` | Most-recent N sessions |
 | `--session <path>` | One explicit session file |
 | `--cwd <dir>` | Filter by working directory |
-| `--since <t>` | `upload`: window — `30m`/`2h`/`7d` or ISO (default 24h); `analyze`: ISO cutoff |
+| `--since <t>` | `upload`: window, `30m`/`2h`/`7d` or ISO (default 24h); `analyze`: ISO cutoff |
 | `--out <path>` | Write the report to a file |
 | `--dir <path>` | `improve`: write the full artifact pack to this directory |
 | `--otlp <path>` | OTLP artifact path (also evidence provenance / dry-run upload preview) |
@@ -130,7 +130,7 @@ traces upload   --since 24h                        # upload last day to the Inte
 | `--mode <kind>` | `stream`: `visualizer` (spans + findings), `findings` (low-volume), or `agent` (findings + reports) |
 | `--replay` / `--once` | `stream`: scan once, then exit |
 | `--no-spans` / `--no-findings` | `stream`: suppress raw span rows / finding rows |
-| `--no-content` | `upload`: send metadata only — strip all prompt/response text |
+| `--no-content` | `upload`: send metadata only; strip all prompt/response text |
 | `--dry-run` / `--yes` | `upload`: preview without sending / skip the confirm prompt |
 
 ## Live stream
@@ -152,6 +152,36 @@ Live streaming emits `session`, `span`, `analysis_batch`, `finding`, and `tick`
 The semantic findings currently cover repeated failing commands, verification churn without code/config changes, completion claims without later verification, and high tool-error rates.
 Use `--mode findings` when you want the low-volume meaning layer; keep `visualizer` for real-time views that need motion, timing, and tool-call texture.
 Use `--mode agent` when another agent needs deterministic loop/tool-use reports alongside the findings.
+The same JSONL shape is used for live sessions and replayed trace files:
+
+![traces stream findings mode terminal capture](./docs/stream-findings.svg)
+
+Custom online analysts use the same config file as `investigate` and `improve`:
+
+```js
+export default {
+  liveAnalysts: [{
+    id: 'my-live-rule',
+    analyze(context) {
+      return context.actions.some((action) => action.kind === 'claim') ? [{
+        schemaVersion: 1,
+        kind: 'traces.live_finding',
+        id: `live.my-live-rule.${context.session.sessionId}`,
+        ruleId: 'my-live-rule',
+        fingerprint: `my-live-rule.${context.session.sessionId}`,
+        severity: 'info',
+        title: 'Completion claim seen',
+        claim: 'A claim-like assistant message appeared in the live trace.',
+        action: 'Require the next stream batch to include a verification action.',
+        check: 'A later finding or batch should show a test/build/typecheck action.',
+        evidence: [{ kind: 'metric', label: 'actions', value: String(context.actions.length) }],
+        session: context.session,
+        observedAt: context.generatedAt,
+      }] : []
+    },
+  }],
+}
+```
 
 ## Improvement engine
 
@@ -229,7 +259,7 @@ That boundary matters. `agent-lab` campaign `cells.jsonl` says "arm X beat arm Y
 
 ### Export existing evidence/events to OpenInference
 
-If you already have compact evidence or Sandbox/OpenCode event captures on disk, convert them to the same OpenInference JSONL shape that `traces analyze --analyzer halo` uses:
+If you already have compact evidence or event captures on disk, convert them to the same OpenInference JSONL shape that `traces analyze --analyzer halo` uses:
 
 ```bash
 traces export policy-evidence.jsonl --out spans.openinference.jsonl
@@ -240,7 +270,7 @@ halo spans.openinference.jsonl --prompt "Analyze this trace slice" --max-turns 1
 `traces export` accepts:
 
 - compact `traces.policy_evidence.session` JSONL from `traces evidence`
-- Sandbox/OpenCode JSON arrays with `start`, `raw`, `result`, `done`, and `error` events
+- JSON arrays with `start`, `raw`, `result`, `done`, and `error` events
 - existing OpenInference JSONL, rewritten through the local redaction path
 
 Run `traces export --help` for the full command reference.
@@ -250,28 +280,28 @@ Run `traces export --help` for the full command reference.
 `upload` **redacts locally before anything leaves the machine**, dedups against already-uploaded sessions, and tags each with metadata (harness, cwd, git branch, host).
 
 ```bash
-traces upload --since 24h --dry-run     # see exactly what would be sent — no network
-traces upload --since 24h --no-content  # send metadata only — drop all prompt/response text
+traces upload --since 24h --dry-run     # see exactly what would be sent; no network
+traces upload --since 24h --no-content  # send metadata only; drop all prompt/response text
 traces upload --since 24h               # send it
 ```
 
 It needs `TANGLE_INGEST_URL` (or `TANGLE_ORCHESTRATOR_URL`), `TANGLE_INGEST_API_KEY` (or `TANGLE_API_KEY`), and `TANGLE_TENANT_ID`. Without them, `--dry-run` still works fully.
 
-### Redaction scope — read this before uploading prose
+### Redaction scope: read this before uploading prose
 
 Redaction is **best-effort regex** for *structured* secrets and credentials: API keys, GitHub/cloud tokens, JWTs, bearer headers, private-key blocks, `KEY=secret` assignments, and credentials embedded in URLs. It runs over every span attribute, including the captured prompt/response text.
 
-It does **not** catch free-form PII — names, postal addresses, phone numbers in prose — which needs a context-aware model. Three postures, strongest first:
+It does **not** catch free-form PII such as names, postal addresses, or phone numbers in prose. Those need a context-aware model. Three postures, strongest first:
 
-1. **`--no-content`** — upload metadata only (tool calls, tokens, timing, loop signal); no prose leaves the machine.
+1. **`--no-content`**: upload metadata only (tool calls, tokens, timing, loop signal); no prose leaves the machine.
 2. Run an ML PII scrubber (e.g. [`openai/privacy-filter`](https://github.com/openai/privacy-filter)) on the platform ingest side as defense-in-depth.
-3. Default — regex redaction of structured secrets.
+3. Default: regex redaction of structured secrets.
 
 Always `--dry-run` first to see exactly what would be sent.
 
 ## External engines (bring your own)
 
-`traces` hosts analysis engines and PII scrubbers it does **not** bundle — you install the tool, `traces` drives it over a thin command adapter. Same pattern for any future engine or model.
+`traces` hosts analysis engines and PII scrubbers it does **not** bundle. You install the tool; `traces` drives it over a thin command adapter. Same pattern for any future engine or model.
 
 **Analyzers** run over the emitted OTLP artifact as peers to the built-in analysts:
 
@@ -281,7 +311,7 @@ traces analyze --last 1 --analyzer halo --analyzer-prompt "find token waste"
 traces analyze --last 1 --analyzer halo --analyzer my-engine    # repeatable
 ```
 
-Our OTLP artifact is **canonical OpenInference** (top-level `kind`, `resource`, `scope`), so HALO reads it directly — no conversion. HALO runs its *own* LLM (OpenAI client — set `OPENAI_BASE_URL` / `OPENAI_API_KEY`, or use HALO's provider); `--model` is forwarded to it. traces supplies the trace and drives the CLI; it doesn't pay for or configure HALO's model.
+Our OTLP artifact is **canonical OpenInference** (top-level `kind`, `resource`, `scope`), so HALO reads it directly with no conversion. HALO runs its *own* LLM. Set `OPENAI_BASE_URL` / `OPENAI_API_KEY`, or use HALO's provider; `--model` is forwarded to it. `traces` supplies the trace and drives the CLI; it does not pay for or configure HALO's model.
 
 **Redactors** scrub prompt/response prose with an external PII model (catching names/addresses the regex pass can't), running *after* the built-in redaction:
 
@@ -292,7 +322,7 @@ traces upload --since 24h --dry-run --redactor "my-pii-scrubber"
 
 In the SDK these are the `ExternalAnalyzer` and `Redactor` interfaces (`haloAnalyzer`, `commandAnalyzer`, `commandRedactor`, `applyRedactor`, `runExternalAnalyzers`). See [`examples/external-engines.ts`](./examples/external-engines.ts).
 
-> For the built-in agentic analysts (`--llm`), set `OPENAI_API_KEY` — or point at any OpenAI-compatible gateway with `OPENAI_BASE_URL` (e.g. an internal router) to use a non-OpenAI key.
+> For the built-in agentic analysts (`--llm`), set `OPENAI_API_KEY`, or point at any OpenAI-compatible gateway with `OPENAI_BASE_URL` (e.g. an internal router) to use a non-OpenAI key.
 
 ## Release automation
 
@@ -312,7 +342,7 @@ The CLI is a thin consumer of these exports.
 
 | Export | Signature | Use |
 |---|---|---|
-| `analyzeSpans` | `(spans, { registry?, ai?, budgetUsd? }) → AnalyzeResult` | run analysts — built-in, or **your own** via `registry` |
+| `analyzeSpans` | `(spans, { registry?, ai?, budgetUsd? }) → AnalyzeResult` | run built-in analysts, or **your own** via `registry` |
 | `runTraceInvestigation` | `(TraceInvestigationOptions) → TraceInvestigationResult` | typed findings, recommendations, claims, external analyzer output, and report |
 | `runTraceImprovementLoop` | `(TraceImprovementOptions) → TraceImprovementResult` | writes the full improvement artifact pack and optional proposal output |
 | `buildTraceFindingPacket` | `({ findings }) → TraceFindingPacket` | turn any `AnalystFinding[]` into recommendations, claims, and a report |
@@ -366,7 +396,7 @@ Runnable, in [`examples/`](./examples):
 |---|---|
 | [`observe-and-alert.ts`](./examples/observe-and-alert.ts) | tail live sessions and alert on stuck loops |
 | [`custom-analyst.ts`](./examples/custom-analyst.ts) | register and run your own analyst |
-| [`improvement-config.mjs`](./examples/improvement-config.mjs) | plug in BYO analysts and proposal generation for `traces improve` |
+| [`improvement-config.mjs`](./examples/improvement-config.mjs) | plug in BYO analysts for `traces stream` / `traces improve` |
 | [`custom-backend.ts`](./examples/custom-backend.ts) | redact + dedup + upload to your own sink |
 | [`register-harness.ts`](./examples/register-harness.ts) | add a new harness by implementing `HarnessTraceAdapter` |
 | [`external-engines.ts`](./examples/external-engines.ts) | drive HALO + an external PII scrubber you install yourself |
@@ -378,5 +408,5 @@ pnpm install
 pnpm dev analyze --harness claude-code --last 1   # run from source via tsx
 pnpm test
 pnpm typecheck
-pnpm build        # → dist/index.js (SDK) + dist/cli.js (bin) + .d.ts
+pnpm build        # dist/index.js (SDK) + dist/cli.js (bin) + .d.ts
 ```

package/dist/cli.js

@@ -215,7 +215,7 @@ async function buildAxService() {
   const apiKey = process.env.OPENAI_API_KEY;
   if (!apiKey) {
     throw new Error(
-      "--llm needs OPENAI_API_KEY \u2014 an OpenAI key, or a router/gateway key with OPENAI_BASE_URL set to an OpenAI-compatible endpoint (e.g. the Tangle router). Deterministic analysis needs no key."
+      "--llm needs OPENAI_API_KEY: an OpenAI key, or a router/gateway key with OPENAI_BASE_URL set to an OpenAI-compatible endpoint (e.g. the Tangle router). Deterministic analysis needs no key."
     );
   }
   const apiURL = process.env.OPENAI_BASE_URL || void 0;
@@ -226,7 +226,7 @@ async function cmdList(args) {
   const groups = await discover({ ...args, last: args.last || 20 });
   for (const { adapter, refs } of groups) {
     console.log(`
-${adapter.harness} \u2014 ${refs.length} session(s)`);
+${adapter.harness}: ${refs.length} session(s)`);
     for (const r of refs) {
       console.log(`  ${new Date(r.mtimeMs).toISOString()}  ${r.sessionId}  ${r.cwd ?? ""}`);
     }
@@ -516,7 +516,7 @@ function formatLiveFinding(finding) {
   const where = `[${finding.session.harness}] ${finding.session.sessionId.slice(0, 8)}`;
   const cwd = finding.session.cwd ? ` \xB7 ${finding.session.cwd}` : "";
   return [
-    `${ts} ${finding.severity.toUpperCase()} ${where} \u2014 ${finding.title}${cwd}`,
+    `${ts} ${finding.severity.toUpperCase()} ${where}: ${finding.title}${cwd}`,
     `  evidence: ${summarizeFindingEvidence(finding)}`,
     `  action: ${finding.action}`,
     `  check: ${finding.check}`
@@ -591,7 +591,7 @@ async function cmdStream(args) {
   process.once("SIGINT", () => controller.abort());
   if (!args.replay) {
     process.stderr.write(
-      `traces stream \u2014 ${preset.mode} JSONL feed for ${all ? "all harnesses" : args.harness}, last ${args.window}m, every ${args.interval}s. Ctrl-C to stop.
+      `traces stream: ${preset.mode} JSONL feed for ${all ? "all harnesses" : args.harness}, last ${args.window}m, every ${args.interval}s. Ctrl-C to stop.
 `
     );
   }
@@ -624,7 +624,7 @@ async function cmdWatch(args) {
   const controller = new AbortController();
   process.once("SIGINT", () => controller.abort());
   process.stderr.write(
-    `traces watch \u2014 observing ${all ? "all harnesses" : args.harness}, sessions active in the last ${args.window}m, every ${args.interval}s. Loop + semantic live findings; read-only; Ctrl-C to stop.
+    `traces watch: observing ${all ? "all harnesses" : args.harness}, sessions active in the last ${args.window}m, every ${args.interval}s. Loop + semantic live findings; read-only; Ctrl-C to stop.
 `
   );
   await watchSessions({
@@ -638,7 +638,7 @@ async function cmdWatch(args) {
     onLoop: (l) => {
       const ts = (/* @__PURE__ */ new Date()).toISOString().slice(11, 19);
       process.stdout.write(
-        `${ts} LOOP [${l.harness}] ${l.sessionId.slice(0, 8)} \u2014 \`${l.toolName}\` repeated \xD7${l.occurrences} with identical args (${(l.windowMs / 1e3).toFixed(0)}s)${l.cwd ? ` \xB7 ${l.cwd}` : ""}
+        `${ts} LOOP [${l.harness}] ${l.sessionId.slice(0, 8)}: \`${l.toolName}\` repeated \xD7${l.occurrences} with identical args (${(l.windowMs / 1e3).toFixed(0)}s)${l.cwd ? ` \xB7 ${l.cwd}` : ""}
 `
       );
     },
@@ -686,7 +686,7 @@ Window: since ${new Date(sinceMs).toISOString()}
   }
   if (args.dryRun) {
     const res2 = await executeUpload(plan, { dryRun: true, otlpOut: args.otlp, stripContent: args.noContent, redactor });
-    console.log(`dry run \u2014 ${newItems.length} session(s), ${totalRedactions} redaction(s). Redacted OTLP \u2192 ${res2.otlpPath}`);
+    console.log(`dry run: ${newItems.length} session(s), ${totalRedactions} redaction(s). Redacted OTLP -> ${res2.otlpPath}`);
     console.log("No upload performed. Set TANGLE_INGEST_URL / TANGLE_INGEST_API_KEY / TANGLE_TENANT_ID and drop --dry-run to send.");
     return;
   }
@@ -704,7 +704,7 @@ Window: since ${new Date(sinceMs).toISOString()}
   );
 }
 function usageExport() {
-  console.log(`traces export \u2014 convert trace evidence files to OpenInference JSONL
+  console.log(`traces export: convert trace evidence files to OpenInference JSONL
 
 Usage:
   traces export <input.jsonl|input.json> --out <spans.jsonl> [--format auto]
@@ -714,7 +714,7 @@ Usage:
 
 Input formats:
   policy-evidence  Compact JSONL rows from \`traces evidence --out policy-evidence.jsonl\`
-  sandbox-events   Sandbox/OpenCode event arrays with start/raw/result/done/error events
+  sandbox-events   JSON arrays with start/raw/result/done/error events
   openinference    Existing OpenInference JSONL; rewrites through traces redaction
   intelligence-spans  JSONL rows exported from Tangle Intelligence trace spans
   auto             Detect the format from the file contents (default)
@@ -732,7 +732,7 @@ Safety:
   --metadata must be a JSON object; --attr key=value is repeatable and overrides matching metadata keys.`);
 }
 function usage() {
-  console.log(`traces \u2014 analyze, observe & upload coding-agent sessions
+  console.log(`traces: analyze, observe & upload coding-agent sessions
 
 Commands:
   list      List discovered sessions
@@ -754,7 +754,7 @@ Options:
   --last <n>       Most-recent N sessions
   --session <path> Analyze/stream one explicit harness session file
   --cwd <dir>      Filter sessions by working directory
-  --since <t>      upload: window \u2014 30m / 2h / 7d or an ISO date (default 24h); analyze: ISO cutoff
+  --since <t>      upload: window, 30m / 2h / 7d or an ISO date (default 24h); analyze: ISO cutoff
   --out <path>     Write report to a file
   --dir <path>     improve: write artifacts to this directory
   --otlp <path>    OTLP artifact path (also evidence provenance / dry-run upload preview)
@@ -771,11 +771,11 @@ Options:
   --budget <usd>   USD cap for agentic analysts
   --analyzer <cmd> analyze: also run an external engine over the OTLP (repeatable; "halo" or any command)
   --analyzer-prompt <p>  analyze: prompt passed to external analyzers (default: diagnose)
-  --interval <s>   watch: poll interval seconds (default 5)
-  --window <m>     watch: only sessions active in the last N minutes (default 30)
+  --interval <s>   watch/stream: poll interval seconds (default 5)
+  --window <m>     watch/stream: only sessions active in the last N minutes (default 30)
   --min-loop <n>   Min identical repeated calls to flag a loop (default 3)
   --dry-run        upload: redact + dedup + preview, write OTLP, but do NOT send
-  --no-content     upload: strip prompt/response text \u2014 send metadata only
+  --no-content     upload: strip prompt/response text; send metadata only
   --redactor <cmd> upload: external PII scrubber (JSON array stdin\u2192stdout) after the regex pass
   --yes, -y        upload: skip the confirmation prompt
   --version, -v    Print the installed traces version

package/docs/stream-findings.svg

@@ -0,0 +1,39 @@
+<svg xmlns="http://www.w3.org/2000/svg" role="img" viewBox="0 0 1100 540">
+  <title>traces stream findings mode terminal capture</title>
+  <desc>A terminal preview of traces stream findings mode showing session, batch, and finding JSONL rows.</desc>
+  <rect width="1100" height="540" fill="#f6f7f9"/>
+  <rect x="50" y="44" width="1000" height="452" rx="8" fill="#111827"/>
+  <rect x="50" y="44" width="1000" height="42" rx="8" fill="#202938"/>
+  <circle cx="78" cy="65" r="7" fill="#ff5f57"/>
+  <circle cx="100" cy="65" r="7" fill="#febc2e"/>
+  <circle cx="122" cy="65" r="7" fill="#28c840"/>
+  <text x="550" y="70" text-anchor="middle" font-family="ui-monospace, SFMono-Regular, Menlo, Consolas, monospace" font-size="14" fill="#cbd5e1">traces stream --mode findings</text>
+
+  <text x="78" y="124" font-family="ui-monospace, SFMono-Regular, Menlo, Consolas, monospace" font-size="18" fill="#8bd5ff">$ traces stream spans.openinference.jsonl --format openinference --mode findings</text>
+
+  <g font-family="ui-monospace, SFMono-Regular, Menlo, Consolas, monospace" font-size="16">
+    <text x="78" y="174" fill="#e5e7eb">{&quot;event&quot;:&quot;session&quot;,</text>
+    <text x="313" y="174" fill="#a7f3d0">&quot;spanCount&quot;:5,</text>
+    <text x="458" y="174" fill="#e5e7eb">&quot;sessionId&quot;:&quot;spans.openinference.jsonl&quot;}</text>
+
+    <text x="78" y="224" fill="#e5e7eb">{&quot;event&quot;:&quot;analysis_batch&quot;,</text>
+    <text x="352" y="224" fill="#a7f3d0">&quot;findingCount&quot;:3,</text>
+    <text x="536" y="224" fill="#fcd34d">&quot;toolCallCount&quot;:3,</text>
+    <text x="732" y="224" fill="#fca5a5">&quot;erroredToolCallCount&quot;:3}</text>
+
+    <text x="78" y="274" fill="#e5e7eb">{&quot;event&quot;:&quot;finding&quot;, &quot;finding&quot;:{</text>
+    <text x="407" y="274" fill="#fca5a5">&quot;severity&quot;:&quot;high&quot;,</text>
+    <text x="608" y="274" fill="#c4b5fd">&quot;ruleId&quot;:&quot;same-failing-command&quot;}}</text>
+    <text x="102" y="304" fill="#94a3b8">action: Stop rerunning it; inspect the first failure before another retry.</text>
+
+    <text x="78" y="354" fill="#e5e7eb">{&quot;event&quot;:&quot;finding&quot;, &quot;finding&quot;:{</text>
+    <text x="407" y="354" fill="#fca5a5">&quot;severity&quot;:&quot;high&quot;,</text>
+    <text x="608" y="354" fill="#c4b5fd">&quot;ruleId&quot;:&quot;verification-without-change&quot;}}</text>
+    <text x="102" y="384" fill="#94a3b8">evidence: 3 verification calls since the last detected code/config change.</text>
+
+    <text x="78" y="434" fill="#e5e7eb">{&quot;event&quot;:&quot;finding&quot;, &quot;finding&quot;:{</text>
+    <text x="407" y="434" fill="#fcd34d">&quot;severity&quot;:&quot;medium&quot;,</text>
+    <text x="627" y="434" fill="#c4b5fd">&quot;ruleId&quot;:&quot;completion-claim-without-verification&quot;}}</text>
+    <text x="102" y="464" fill="#94a3b8">check: a later test/build/typecheck/lint/smoke command should appear.</text>
+  </g>
+</svg>

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tangle-network/traces",
-  "version": "0.8.19",
+  "version": "0.8.21",
   "description": "Point it at your coding-agent session traces (Claude Code, Codex, OpenCode, Gemini, Pi, …) and get failure-mode + efficiency findings. CLI + SDK over the @tangle-network/agent-eval analyst suite — observe live sessions, run your own analysts, redact, and upload to the Tangle Intelligence Platform.",
   "type": "module",
   "license": "MIT",
@@ -36,6 +36,7 @@
   },
   "files": [
     "dist",
+    "docs/stream-findings.svg",
     "install.sh",
     "README.md"
   ],