pi-baseten-provider 1.0.9 → 1.0.11

package/README.md

@@ -71,18 +71,17 @@ pi
 |-------|---------|--------|-----------|-----------|------------|
 | Deepseek V4 Flash 0731 | 1.0M | ❌ | ✅ | $0.13 | $0.26 |
 | Deepseek V4 Pro | 262K | ❌ | ✅ | $1.74 | $3.48 |
+| Deepseek V4 Pro 0813 | 1.0M | ❌ | ✅ | $1.32 | $3.96 |
 | GLM 4.7 | 200K | ❌ | ✅ | $0.60 | $2.20 |
 | GLM 5.2 | 1.0M | ❌ | ✅ | $1.40 | $4.40 |
-| GLM 5.2 Fast | 524K | ❌ | ✅ | $2.10 | $6.60 |
+| GLM 5.2 Fast | 1.0M | ❌ | ✅ | $2.10 | $6.60 |
 | Inkling | 1.0M | ✅ | ✅ | $1.00 | $4.05 |
 | Inkling Small | 1.0M | ✅ | ✅ | $0.50 | $1.20 |
 | Kimi K2.6 | 262K | ✅ | ✅ | $0.95 | $4.00 |
 | Kimi K2.7 Code | 262K | ✅ | ✅ | $0.95 | $4.00 |
 | Kimi K3 | 1.0M | ✅ | ✅ | $3.00 | $15.00 |
-| Mercury 2 | 8K | ❌ | ✅ | Free | Free |
 | Nemotron Ultra | 203K | ❌ | ✅ | $0.60 | $2.40 |
 | OpenAI GPT 120B | 128K | ❌ | ✅ | $0.10 | $0.50 |
-| SID-1 | 33K | ❌ | ❌ | Free | Free |
 
 *Costs are per million tokens. Prices subject to change — check [baseten.co/pricing](https://www.baseten.co/pricing/) for current pricing.*
 

package/deprecated-models.json

@@ -1 +1,48 @@
-{}
+{
+  "inception/mercury-2": {
+    "id": "inception/mercury-2",
+    "name": "Mercury 2",
+    "reasoning": true,
+    "input": [
+      "text"
+    ],
+    "cost": {
+      "input": 0,
+      "output": 0,
+      "cacheRead": 0,
+      "cacheWrite": 0
+    },
+    "contextWindow": 8192,
+    "maxTokens": 5000,
+    "compat": {
+      "supportsDeveloperRole": true,
+      "supportsStore": false,
+      "maxTokensField": "max_completion_tokens",
+      "thinkingFormat": "openai"
+    },
+    "deprecatedAt": "2026-08-05T02:00:12.114Z"
+  },
+  "sid/sid-1": {
+    "id": "sid/sid-1",
+    "name": "SID-1",
+    "reasoning": true,
+    "input": [
+      "text"
+    ],
+    "cost": {
+      "input": 0,
+      "output": 0,
+      "cacheRead": 0,
+      "cacheWrite": 0
+    },
+    "contextWindow": 32768,
+    "maxTokens": 5000,
+    "compat": {
+      "supportsDeveloperRole": true,
+      "supportsStore": false,
+      "maxTokensField": "max_completion_tokens",
+      "thinkingFormat": "openai"
+    },
+    "deprecatedAt": "2026-08-05T02:00:12.114Z"
+  }
+}

package/models.json

@@ -43,6 +43,28 @@
       "thinkingFormat": "openai"
     }
   },
+  {
+    "id": "deepseek-ai/DeepSeek-V4-Pro-0813",
+    "name": "Deepseek V4 Pro 0813",
+    "reasoning": true,
+    "input": [
+      "text"
+    ],
+    "cost": {
+      "input": 1.32,
+      "output": 3.96,
+      "cacheRead": 0.132,
+      "cacheWrite": 0
+    },
+    "contextWindow": 1000000,
+    "maxTokens": 262144,
+    "compat": {
+      "supportsDeveloperRole": true,
+      "supportsStore": false,
+      "maxTokensField": "max_completion_tokens",
+      "thinkingFormat": "openai"
+    }
+  },
   {
     "id": "zai-org/GLM-4.7",
     "name": "GLM 4.7",
@@ -100,7 +122,7 @@
       "cacheRead": 0.21,
       "cacheWrite": 0
     },
-    "contextWindow": 524288,
+    "contextWindow": 1048576,
     "maxTokens": 262144,
     "compat": {
       "supportsDeveloperRole": true,
@@ -224,28 +246,6 @@
       "thinkingFormat": "openai"
     }
   },
-  {
-    "id": "inception/mercury-2",
-    "name": "Mercury 2",
-    "reasoning": true,
-    "input": [
-      "text"
-    ],
-    "cost": {
-      "input": 0,
-      "output": 0,
-      "cacheRead": 0,
-      "cacheWrite": 0
-    },
-    "contextWindow": 8192,
-    "maxTokens": 5000,
-    "compat": {
-      "supportsDeveloperRole": true,
-      "supportsStore": false,
-      "maxTokensField": "max_completion_tokens",
-      "thinkingFormat": "openai"
-    }
-  },
   {
     "id": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B",
     "name": "Nemotron Ultra",
@@ -290,27 +290,5 @@
       "thinkingFormat": "openai",
       "supportsReasoningEffort": true
     }
-  },
-  {
-    "id": "sid/sid-1",
-    "name": "SID-1",
-    "reasoning": true,
-    "input": [
-      "text"
-    ],
-    "cost": {
-      "input": 0,
-      "output": 0,
-      "cacheRead": 0,
-      "cacheWrite": 0
-    },
-    "contextWindow": 32768,
-    "maxTokens": 5000,
-    "compat": {
-      "supportsDeveloperRole": true,
-      "supportsStore": false,
-      "maxTokensField": "max_completion_tokens",
-      "thinkingFormat": "openai"
-    }
   }
 ]

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "pi-baseten-provider",
-  "version": "1.0.9",
+  "version": "1.0.11",
   "description": "Baseten provider extension for pi - Access DeepSeek, Kimi, GLM, MiniMax, Nemotron, and GPT-OSS models through the Baseten Model API",
   "author": "monotykamary",
   "homepage": "https://github.com/monotykamary/pi-baseten-provider#readme",

package/patch.json

@@ -1,5 +1,6 @@
 {
   "deepseek-ai/DeepSeek-V4-Flash-0731": {
+    "maxTokens": 384000,
     "thinkingLevelMap": {
       "off": "none",
       "minimal": null,

package/scripts/update-models.js

@@ -17,15 +17,119 @@
  * patch.json and custom-models.json are applied at runtime by the provider.
  * They are NOT baked into models.json, but ARE used to generate the README table.
  *
- * Requires BASETEN_API_KEY environment variable.
+ * API key: the stored `baseten` credential in ~/.pi/agent/auth.json wins, then
+ * the BASETEN_API_KEY environment variable. The script refuses to run without one.
  */
 
 import fs from 'fs';
+import os from 'os';
+import { execSync } from 'child_process';
 import path from 'path';
 import { fileURLToPath } from 'url';
 
 const __dirname = path.dirname(fileURLToPath(import.meta.url));
 
+// pi's agent directory: PI_CODING_AGENT_DIR (with ~ expansion) or ~/.pi/agent.
+function piAgentDir() {
+  const envDir = process.env.PI_CODING_AGENT_DIR;
+  if (envDir) {
+    return envDir.startsWith('~/') || envDir === '~'
+      ? path.join(os.homedir(), envDir.slice(1))
+      : envDir;
+  }
+  return path.join(os.homedir(), '.pi', 'agent');
+}
+
+const AUTH_JSON_PATH = path.join(piAgentDir(), 'auth.json');
+
+/**
+ * Resolve a configured value using pi's semantics (resolve-config-value.ts in
+ * pi-mono): "!command" runs via the shell (10s timeout) and uses trimmed
+ * stdout; "$VAR" / "${VAR}" interpolate environment variables ("$$" escapes a
+ * literal "$", "$!" a literal "!"); anything else is a literal. Returns
+ * undefined when a referenced env var is unset or a command fails.
+ */
+function resolveConfigValue(config, env) {
+  if (typeof config !== 'string' || config.length === 0) return undefined;
+  if (config.startsWith('!')) {
+    try {
+      const out = execSync(config.slice(1), {
+        encoding: 'utf8',
+        timeout: 10000,
+        stdio: ['ignore', 'pipe', 'ignore'],
+      });
+      return out.trim() || undefined;
+    } catch {
+      return undefined;
+    }
+  }
+  const ENV_NAME_RE = /^[A-Za-z_][A-Za-z0-9_]*$/;
+  let resolved = '';
+  let index = 0;
+  while (index < config.length) {
+    const dollar = config.indexOf('$', index);
+    if (dollar < 0) {
+      resolved += config.slice(index);
+      break;
+    }
+    resolved += config.slice(index, dollar);
+    const next = config[dollar + 1];
+    let name;
+    if (next === '$' || next === '!') {
+      resolved += next;
+      index = dollar + 2;
+      continue;
+    } else if (next === '{') {
+      const end = config.indexOf('}', dollar + 2);
+      if (end < 0) {
+        resolved += '$';
+        index = dollar + 1;
+        continue;
+      }
+      const inner = config.slice(dollar + 2, end);
+      if (!ENV_NAME_RE.test(inner)) {
+        resolved += config.slice(dollar, end + 1);
+        index = end + 1;
+        continue;
+      }
+      name = inner;
+      index = end + 1;
+    } else {
+      const match = config.slice(dollar + 1).match(/^[A-Za-z_][A-Za-z0-9_]*/);
+      if (!match) {
+        resolved += '$';
+        index = dollar + 1;
+        continue;
+      }
+      name = match[0];
+      index = dollar + 1 + name.length;
+    }
+    const value = (env && env[name]) || process.env[name] || undefined;
+    if (value === undefined) return undefined;
+    resolved += value;
+  }
+  return resolved;
+}
+
+/**
+ * The API key, resolved the way pi itself resolves it for this provider: the
+ * stored `baseten` credential in ~/.pi/agent/auth.json wins, then
+ * the BASETEN_API_KEY environment variable.
+ */
+function resolveApiKey() {
+  try {
+    const auth = JSON.parse(fs.readFileSync(AUTH_JSON_PATH, 'utf8'));
+    const credential = auth?.baseten;
+    if (credential && credential.type === 'api_key' && typeof credential.key === 'string') {
+      const key = resolveConfigValue(credential.key, credential.env);
+      if (key) return key;
+    }
+  } catch {
+    // Missing or unparseable auth.json: fall through to the env var.
+  }
+  return process.env.BASETEN_API_KEY || undefined;
+}
+
 const MODELS_API_URL = 'https://inference.baseten.co/v1/models';
 const MODELS_JSON_PATH = path.join(__dirname, '..', 'models.json');
 const PATCH_JSON_PATH = path.join(__dirname, '..', 'patch.json');
@@ -58,9 +162,9 @@ function toPerMillion(val) {
 // ─── API fetch ───────────────────────────────────────────────────────────────
 
 async function fetchModels() {
-  const apiKey = process.env.BASETEN_API_KEY;
+  const apiKey = resolveApiKey();
   if (!apiKey) {
-    throw new Error('BASETEN_API_KEY environment variable is required');
+    throw new Error('No API key found: no `baseten` credential resolved from ' + AUTH_JSON_PATH + ' and BASETEN_API_KEY is not set');
   }
 
   console.log(`Fetching models from ${MODELS_API_URL}...`);