chore(deps): accept the published first-party cohort

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tangle-network/braid",
-  "version": "0.2.0",
+  "version": "0.2.1",
   "description": "A universal terminal interface for portable agent profiles",
   "type": "module",
   "bin": {
@@ -100,11 +100,11 @@
     "@dataiku/uv": "0.12.0",
     "@earendil-works/pi-tui": "0.84.2",
     "@napi-rs/keyring": "1.3.0",
-    "@tangle-network/agent-eval": "0.149.0",
-    "@tangle-network/agent-interface": "1.3.0",
-    "@tangle-network/agent-provider-cli-bridge": "0.9.4",
-    "@tangle-network/agent-provider-tangle": "0.13.0",
-    "@tangle-network/agent-runtime": "0.143.0",
+    "@tangle-network/agent-eval": "0.163.2",
+    "@tangle-network/agent-interface": "1.4.0",
+    "@tangle-network/agent-provider-cli-bridge": "0.9.5",
+    "@tangle-network/agent-provider-tangle": "0.13.1",
+    "@tangle-network/agent-runtime": "0.153.2",
     "@tangle-network/sandbox": "0.31.0",
     "better-sqlite3-multiple-ciphers": "13.0.3",
     "chalk": "6.0.0",

pnpm-workspace.yaml

@@ -9,13 +9,13 @@ allowBuilds:
   node-pty: true
 ignoreScripts: false
 minimumReleaseAgeExclude:
-  - '@tangle-network/agent-runtime@0.126.0 || 0.128.0 || 0.131.0 || 0.131.1 || 0.131.2 || 0.131.3 || 0.131.4 || 0.131.5 || 0.131.6 || 0.131.7 || 0.132.0 || 0.132.4 || 0.132.6 || 0.132.9 || 0.132.10 || 0.132.11 || 0.132.12 || 0.133.6 || 0.134.2 || 0.134.4 || 0.134.5 || 0.134.9 || 0.135.0 || 0.135.3 || 0.142.0 || 0.142.1 || 0.142.3 || 0.143.0'
-  - '@tangle-network/agent-eval@0.143.0 || 0.144.0 || 0.144.1 || 0.144.3 || 0.144.4 || 0.144.6 || 0.144.8 || 0.144.10 || 0.144.11 || 0.144.13 || 0.145.0 || 0.145.1 || 0.145.2 || 0.145.3 || 0.145.9 || 0.145.10 || 0.145.11 || 0.145.14 || 0.145.15 || 0.145.21 || 0.149.0'
-  - '@tangle-network/agent-interface@0.46.0 || 0.46.1 || 0.47.0 || 0.51.0 || 0.52.0 || 0.53.0 || 0.54.0 || 1.0.0 || 1.1.0 || 1.3.0'
-  - '@tangle-network/agent-knowledge@7.0.8 || 7.1.2 || 7.1.3 || 7.2.0 || 7.2.2 || 7.2.3 || 7.2.4 || 7.2.6 || 8.0.0 || 8.0.1 || 8.0.5 || 8.0.10'
-  - '@tangle-network/agent-profile-materialize@0.10.2 || 0.13.1 || 0.14.0 || 0.14.2 || 0.15.1 || 0.15.2'
-  - '@tangle-network/agent-provider-cli-bridge@0.4.2 || 0.4.3 || 0.5.0 || 0.6.0 || 0.6.4 || 0.7.1 || 0.7.2 || 0.7.3 || 0.7.6 || 0.8.0 || 0.8.1 || 0.9.0 || 0.9.2 || 0.9.3 || 0.9.4'
-  - '@tangle-network/agent-provider-tangle@0.13.0'
+  - '@tangle-network/agent-runtime@0.126.0 || 0.128.0 || 0.131.0 || 0.131.1 || 0.131.2 || 0.131.3 || 0.131.4 || 0.131.5 || 0.131.6 || 0.131.7 || 0.132.0 || 0.132.4 || 0.132.6 || 0.132.9 || 0.132.10 || 0.132.11 || 0.132.12 || 0.133.6 || 0.134.2 || 0.134.4 || 0.134.5 || 0.134.9 || 0.135.0 || 0.135.3 || 0.142.0 || 0.142.1 || 0.142.3 || 0.143.0 || 0.153.2'
+  - '@tangle-network/agent-eval@0.143.0 || 0.144.0 || 0.144.1 || 0.144.3 || 0.144.4 || 0.144.6 || 0.144.8 || 0.144.10 || 0.144.11 || 0.144.13 || 0.145.0 || 0.145.1 || 0.145.2 || 0.145.3 || 0.145.9 || 0.145.10 || 0.145.11 || 0.145.14 || 0.145.15 || 0.145.21 || 0.149.0 || 0.163.2'
+  - '@tangle-network/agent-interface@0.46.0 || 0.46.1 || 0.47.0 || 0.51.0 || 0.52.0 || 0.53.0 || 0.54.0 || 1.0.0 || 1.1.0 || 1.3.0 || 1.4.0'
+  - '@tangle-network/agent-knowledge@7.0.8 || 7.1.2 || 7.1.3 || 7.2.0 || 7.2.2 || 7.2.3 || 7.2.4 || 7.2.6 || 8.0.0 || 8.0.1 || 8.0.5 || 8.0.10 || 10.7.0'
+  - '@tangle-network/agent-profile-materialize@0.10.2 || 0.13.1 || 0.14.0 || 0.14.2 || 0.15.1 || 0.15.2 || 0.17.1'
+  - '@tangle-network/agent-provider-cli-bridge@0.4.2 || 0.4.3 || 0.5.0 || 0.6.0 || 0.6.4 || 0.7.1 || 0.7.2 || 0.7.3 || 0.7.6 || 0.8.0 || 0.8.1 || 0.9.0 || 0.9.2 || 0.9.3 || 0.9.4 || 0.9.5'
+  - '@tangle-network/agent-provider-tangle@0.13.0 || 0.13.1'
   - '@tangle-network/sandbox@0.30.1 || 0.31.0'
   - hono@4.13.0
   - '@esbuild/aix-ppc64@0.28.2'

src/eval/execution.ts

@@ -205,9 +205,14 @@ export function evalJudgeProfile(config: EvalRouteConfig): AgentProfile {
       provider: 'tangle-router',
       default: config.model,
       reasoningEffort: 'none',
+      // One budget bounds the visible answer and the whole completion alike: the judge
+      // returns a short verdict and must never spend the run on reasoning. Runtime lowers
+      // these onto the Router route as `max_tokens` and `max_completion_tokens`, which this
+      // profile used to spell out by hand in `extraBody` and a request-level cap.
+      maxVisibleOutputTokens: EVAL_TOTAL_COMPLETION_TOKENS,
+      maxTotalOutputTokens: EVAL_TOTAL_COMPLETION_TOKENS,
       metadata: {
         temperature: 0,
-        maxTokens: EVAL_TOTAL_COMPLETION_TOKENS,
         retry: {
           maxAttempts: 3,
           initialBackoffMs: 1_000,
@@ -216,8 +221,6 @@ export function evalJudgeProfile(config: EvalRouteConfig): AgentProfile {
           requestTimeoutMs: config.timeoutMs,
         },
         extraBody: {
-          // Router treats this field as the hard ceiling over visible and reasoning tokens.
-          max_completion_tokens: EVAL_TOTAL_COMPLETION_TOKENS,
           // GLM enables thinking by default and does not honor reasoning_effort: none.
           ...(disablesThinking ? { thinking: { type: 'disabled' } } : {}),
         },

test/eval.test.ts

@@ -183,9 +183,10 @@ test('semantic judge execution is one exact AgentProfile owned by Runtime', asyn
       provider: 'tangle-router',
       default: DEFAULT_EVAL_MODEL,
       reasoningEffort: 'none',
+      maxVisibleOutputTokens: EVAL_TOTAL_COMPLETION_TOKENS,
+      maxTotalOutputTokens: EVAL_TOTAL_COMPLETION_TOKENS,
       metadata: {
         temperature: 0,
-        maxTokens: EVAL_TOTAL_COMPLETION_TOKENS,
         retry: {
           maxAttempts: 3,
           initialBackoffMs: 1_000,
@@ -194,7 +195,6 @@ test('semantic judge execution is one exact AgentProfile owned by Runtime', asyn
           requestTimeoutMs: 120_000,
         },
         extraBody: {
-          max_completion_tokens: EVAL_TOTAL_COMPLETION_TOKENS,
           thinking: { type: 'disabled' },
         },
       },
@@ -261,9 +261,9 @@ test('semantic judge execution is one exact AgentProfile owned by Runtime', asyn
   })
 
   const nonGlm = evalJudgeProfile({ ...config, model: 'openai-codex/gpt-5.6-luna' })
-  assert.deepEqual(nonGlm.model?.metadata?.extraBody, {
-    max_completion_tokens: EVAL_TOTAL_COMPLETION_TOKENS,
-  })
+  assert.deepEqual(nonGlm.model?.metadata?.extraBody, {})
+  assert.equal(nonGlm.model?.maxVisibleOutputTokens, EVAL_TOTAL_COMPLETION_TOKENS)
+  assert.equal(nonGlm.model?.maxTotalOutputTokens, EVAL_TOTAL_COMPLETION_TOKENS)
 })
 
 test('semantic judge retries transient Router failures with the same operation identity', async (context) => {
@@ -494,7 +494,7 @@ test('semantic judge refuses an ambiguous smaller request cap before provider sp
       messages: [{ role: 'user', content: 'candidate' }],
       maxTokens: 320,
     }),
-    /request maxTokens conflicts with AgentProfile model metadata/u,
+    /request maxTokens 320 conflicts with AgentProfile\.model\.maxVisibleOutputTokens 2048/u,
   )
   assert.equal(calls, 0)
 })

test/trace-analysis-configuration.test.ts

@@ -587,7 +587,11 @@ test('runtime-owned trace model call preserves canonical messages, limits, usage
       content: JSON.stringify({ messages: optimizerRequest().request.messages }),
     },
   ])
-  assert.equal(receivedBody?.max_tokens, undefined)
+  // The request's cap is the visible ceiling, and Runtime now lowers it onto the Router
+  // route as `max_tokens`; it used to accept the cap and never send it. This profile
+  // declares no total ceiling, so no `max_completion_tokens` is sent.
+  assert.equal(receivedBody?.max_tokens, 64)
+  assert.equal(receivedBody?.max_completion_tokens, undefined)
   assert.equal(receivedBody?.temperature, 0.2)
   assert.deepEqual(receivedBody?.response_format, {
     type: 'json_schema',