martin-loop - npm Package Compare versions

Comparing version 0.3.18 to 0.3.19

dist/vendor/adapters/claude-cli.d.ts

@@ -14,3 +14,3 @@ /**
  */
-import type { MartinAdapter, MartinAdapterRequest } from "../core/index.js";
+import type { MartinAdapter } from "../core/index.js";
 import { type SpawnLike } from "./cli-bridge.js";
@@ -22,3 +22,3 @@ /**
  */
-export type CliArgsBuilder = (prompt: string, request: MartinAdapterRequest) => string[];
+export type CliArgsBuilder = (prompt: string) => string[];
 export type CliStdinBuilder = (prompt: string) => string | undefined;
@@ -25,0 +25,0 @@ export interface AgentCliAdapterOptions {

dist/vendor/adapters/claude-cli.js

@@ -242,3 +242,3 @@ /**
 }
-function createStreamingUsageInspector(capUsd, modelLabel, promptTokenEstimate) {
+function createStreamingUsageInspector(capUsd, modelLabel) {
     const pricing = (modelLabel ? MODEL_PRICING[modelLabel] : undefined) ??
@@ -249,5 +249,3 @@ { inputPer1K: BLENDED_INPUT_COST_PER_1K, outputPer1K: BLENDED_OUTPUT_COST_PER_1K };
     // next check fires (proven live: $1.50 cap → $28.42 actual).
-    const largeContext = promptTokenEstimate > 10_000;
-    const effectiveCapRatio = largeContext ? 0.7 : 0.8;
-    const effectiveCapUsd = capUsd * effectiveCapRatio;
+    const effectiveCapUsd = capUsd * 0.8;
     // Token-count ceiling fallback: if no usage events are ever parsed (e.g.
@@ -276,3 +274,3 @@ // Claude changes its stream-json event format), use raw byte volume as a
             terminate(`Streaming usage cap exceeded after ${String(turns)} turn(s): cumulative cost ~$${cumulativeUsd.toFixed(4)} ` +
-                `surpassed the per-attempt cap $${capUsd.toFixed(4)} (${String(Math.round(effectiveCapRatio * 100))}% threshold: $${effectiveCapUsd.toFixed(4)}). ` +
+                `surpassed the per-attempt cap $${capUsd.toFixed(4)} (80% threshold: $${effectiveCapUsd.toFixed(4)}). ` +
                 `Subprocess terminated to bound runaway overspend.`);
@@ -308,14 +306,2 @@ }
         }
-        const turnUsd = (turnTokensIn / 1000) * pricing.inputPer1K + (turnTokensOut / 1000) * pricing.outputPer1K;
-        const remainingBudgetBeforeTurn = Math.max(capUsd - cumulativeUsd, 0);
-        if (capUsd > 0 && remainingBudgetBeforeTurn > 0 && turnUsd > remainingBudgetBeforeTurn * 0.5) {
-            cumulativeUsd += turnUsd;
-            tokensIn += turnTokensIn;
-            tokensOut += turnTokensOut;
-            turns += 1;
-            usageEventSeen = true;
-            terminate(`Single turn spend ~$${turnUsd.toFixed(4)} consumed more than 50% of the remaining per-attempt budget ` +
-                `($${remainingBudgetBeforeTurn.toFixed(4)} before the turn). Subprocess terminated to prevent a one-turn overshoot.`);
-            return;
-        }
         tokensIn += turnTokensIn;
@@ -325,3 +311,3 @@ tokensOut += turnTokensOut;
         usageEventSeen = true;
-        cumulativeUsd += turnUsd;
+        cumulativeUsd += (turnTokensIn / 1000) * pricing.inputPer1K + (turnTokensOut / 1000) * pricing.outputPer1K;
         checkBudgetExceeded(terminate);
@@ -485,3 +471,3 @@ };
             }
-            const args = options.argsBuilder(prompt, request);
+            const args = options.argsBuilder(prompt);
             const stdinData = options.stdinBuilder?.(prompt);
@@ -495,3 +481,3 @@ // Live cumulative-cost circuit breaker: a single attempt should never be
             const streamingUsage = options.streamingUsageCap && request.context.remainingBudgetUsd > 0
-                ? createStreamingUsageInspector(request.context.remainingBudgetUsd, options.model ?? options.command, estimatedUsage.tokensIn)
+                ? createStreamingUsageInspector(request.context.remainingBudgetUsd, options.model ?? options.command)
                 : undefined;
@@ -814,3 +800,3 @@ const agentResult = await runSubprocess(options.command, args, {
         spawnImpl: options.spawnImpl,
-        argsBuilder: (_prompt, request) => [
+        argsBuilder: (_prompt) => [
             "--output-format",
@@ -826,3 +812,2 @@ "stream-json",
             "--strict-mcp-config",
-            ...(request.context.remainingTokens > 0 ? ["--max-tokens", String(request.context.remainingTokens)] : []),
             ...modelArgs,
@@ -829,0 +814,0 @@ ...extraArgs

dist/vendor/cli/index.d.ts

@@ -138,2 +138,4 @@ import { type MartinAdapter } from "../core/index.js";
     fileScope: string[];
+    cwd?: string;
+    runsDir?: string;
 };
@@ -140,0 +142,0 @@ type GateCommand = {

dist/vendor/cli/package.json

 {
   "name": "@martin/cli",
-  "version": "0.3.18",
+  "version": "0.3.19",
   "type": "module",
@@ -5,0 +5,0 @@ "description": "Martin Loop CLI — budget-aware coding loops with failure classification and verified exits.",

dist/vendor/cli/phase-command-center.js

 import { readFile, readdir, stat } from "node:fs/promises";
 import { join, resolve } from "node:path";
-import { diagnoseCodexHost, resolveCliCommandAvailability } from "../adapters/index.js";
+import { probeCodexLaunch, resolveCliCommandAvailability } from "../adapters/index.js";
 import { resolveRunsRoot } from "../core/index.js";
@@ -400,6 +400,8 @@ const DEFAULT_BLOCKED_PATHS = [
     const availability = resolveCliCommandAvailability("codex");
-    const diagnosis = diagnoseCodexHost(availability);
-    const summary = availability.available
-        ? "Codex CLI detected. Run martin preflight for a live launch check before governed execution."
-        : availability.detail;
+    const probe = mode === "live"
+        ? probeCodexLaunch({
+            workingDirectory: receiptScope.workingDirectory,
+            availability
+        })
+        : undefined;
     return {
@@ -413,10 +415,19 @@ mode,
             ...(availability.candidatePaths?.length ? { candidatePaths: availability.candidatePaths } : {}),
-            hostPlatform: diagnosis.hostPlatform,
-            installKind: diagnosis.installKind,
-            nativeInstallValid: diagnosis.nativeInstallValid,
-            invocationMode: diagnosis.invocationMode,
-            sandboxMode: diagnosis.sandboxMode,
-            sandboxCompatible: diagnosis.sandboxCompatible,
-            summary,
-            ...(diagnosis.remediation ? { remediation: diagnosis.remediation } : {})
+            ...(probe
+                ? {
+                    selectedPath: probe.command,
+                    hostPlatform: probe.diagnosis.hostPlatform,
+                    installKind: probe.diagnosis.installKind,
+                    nativeInstallValid: probe.diagnosis.nativeInstallValid,
+                    invocationMode: probe.diagnosis.invocationMode,
+                    sandboxMode: probe.diagnosis.sandboxMode,
+                    sandboxCompatible: probe.diagnosis.sandboxCompatible,
+                    launchReady: probe.ok,
+                    summary: probe.summary,
+                    ...(probe.diagnosis.remediation ? { remediation: probe.diagnosis.remediation } : {}),
+                    ...(probe.candidateProbeResults?.length
+                        ? { candidateProbeResults: probe.candidateProbeResults }
+                        : {})
+                }
+                : {})
         }
@@ -423,0 +434,0 @@ };

package.json

 {
   "name": "martin-loop",
   "private": false,
-  "version": "0.3.18",
+  "version": "0.3.19",
   "type": "module",
@@ -115,3 +115,15 @@ "description": "Open-source command center for governed AI coding agents with built-in onboarding, hard gates, MCP, and shareable run receipts.",
     "benchmarks"
-  ]
+  ],
+  "pnpm": {
+    "overrides": {
+      "@hono/node-server": "^1.19.13",
+      "ajv": "^8.20.0",
+      "fast-uri": "^3.1.2",
+      "hono": "^4.12.21",
+      "ip-address": "^10.1.1",
+      "postcss": "^8.5.15",
+      "qs": "^6.15.2",
+      "vite": "^7.3.2"
+    }
+  }
 }

README.md

@@ -89,3 +89,3 @@ # MartinLoop
 
-Release notes for the current root package: [MartinLoop 0.3.18](./docs/release/OSS-0.3.18-RELEASE-NOTES.md).
+Release notes for the current root package: [MartinLoop 0.3.19](./docs/release/OSS-0.3.19-RELEASE-NOTES.md).
 
@@ -131,13 +131,13 @@ ## Visual Proof
 ```sh
-npx -y martin-loop@0.3.18 --version
-npx -y martin-loop@0.3.18 start
-npx -y martin-loop@0.3.18 demo
+npx -y martin-loop@0.3.19 --version
+npx -y martin-loop@0.3.19 start
+npx -y martin-loop@0.3.19 demo
 cd martin-loop-demo
 npm install
-npx -y martin-loop@0.3.18 run "Summarize the demo workspace and prove tests still pass" --verify "npm test" --budget-usd 2 --max-iterations 1 --json
-npx -y martin-loop@0.3.18 dossier --latest --json
-npx -y martin-loop@0.3.18 share --latest --json
+npx -y martin-loop@0.3.19 run "Summarize the demo workspace and prove tests still pass" --verify "npm test" --budget-usd 2 --max-iterations 1 --json
+npx -y martin-loop@0.3.19 dossier --latest --json
+npx -y martin-loop@0.3.19 share --latest --json
 ```
 
-For deterministic installs, pin the package line (`martin-loop@0.3.18`) or use `martin-loop@latest`. Plain `npx martin-loop` can resolve a stale local cache on some machines.
+For deterministic installs, pin the package line (`martin-loop@0.3.19`) or use `martin-loop@latest`. Plain `npx martin-loop` can resolve a stale local cache on some machines.
 
@@ -173,5 +173,5 @@ Expected share bundle outputs:
 
-Public governed runs use one canonical taxonomy: the 12 runtime `FailureClass` values from `@martin/contracts`.
+Public governed runs use one canonical taxonomy: the 13 runtime `FailureClass` values from `@martin/contracts`.
 
-See the canonical table: [Failure Taxonomy (13 Runtime Classes)](./docs/oss/FAILURE-TAXONOMY.md).
+See the canonical table: [Failure Taxonomy (13 Runtime Classes)](./docs/oss/FAILURE-TAXONOMY-13.md).
 
@@ -183,3 +183,3 @@ ## What It Does
 - Policy checks block unsafe verifier commands, risky path changes, and secret-like task inputs before execution.
-- Failure classification uses canonical runtime classes for triage and reporting. See [Failure Taxonomy (13 Runtime Classes)](./docs/oss/FAILURE-TAXONOMY.md).
+- Failure classification uses canonical runtime classes for triage and reporting. See [Failure Taxonomy (13 Runtime Classes)](./docs/oss/FAILURE-TAXONOMY-13.md).
 - Run receipts capture stop reason, verifier evidence, budget posture, integrity state, and the next safe action.
@@ -299,8 +299,2 @@ - `martin share --latest` turns the latest governed run into a local share bundle with a redacted JSON receipt, Markdown recap, and proof-card SVG.
 
-If you need a local HTTP endpoint for a bridge or proxy, the standalone package can also run over HTTP:
-
-```sh
-npx -y @martinloop/mcp --http --port 3033
-```
-
 Generate host config from the root CLI:
@@ -315,3 +309,3 @@
 
-The root `martin-loop` package and the standalone `@martinloop/mcp` package move on separate version lines. The root package line here is `0.3.18`; the current standalone MCP package is `0.3.6`.
+The root `martin-loop` package and the standalone `@martinloop/mcp` package move on separate version lines. The current root package line here is `0.3.19`; the current standalone MCP package is `0.3.6`.
 
@@ -375,3 +369,3 @@ The public MCP release train labels are:
 - [Agent Failure Atlas](./docs/agent-failure-atlas.md)
-- [Failure Taxonomy (13 Runtime Classes)](./docs/oss/FAILURE-TAXONOMY.md)
+- [Failure Taxonomy (13 Runtime Classes)](./docs/oss/FAILURE-TAXONOMY-13.md)
 - [PRE-028-PUBLIC-SURFACE-DIFF.md](./docs/oss/PRE-028-PUBLIC-SURFACE-DIFF.md)
@@ -378,0 +372,0 @@ - [Claude Code walkthrough](./docs/getting-started/claude-code.md)

New alerts

AI-detected potential code anomaly

Supply chain risk

AI has identified unusual behaviors that may pose a security risk.

URL strings

Supply chain risk

Package contains fragments of external URLs or IP addresses, which the package may be accessing at runtime.

Fixed alerts

AI-detected potential code anomaly

Supply chain risk

AI has identified unusual behaviors that may pose a security risk.

URL strings

Supply chain risk

Package contains fragments of external URLs or IP addresses, which the package may be accessing at runtime.

Improved metrics

Lines of code

20062

0

Worsened metrics

Total package byte prevSize

875246

0

Number of lines in readme file

437

-1.35%

No dependency changes