@miller-tech/uap 1.26.6 → 1.28.0

package/dist/models/openai-compat-client.d.ts

@@ -0,0 +1,34 @@
+/**
+ * OpenAI-compatible ModelClient
+ *
+ * Production implementation of the ModelClient interface from executor.ts.
+ * Talks to any /v1/chat/completions server — the local inference gateway
+ * (:4000), llama.cpp (:8080), vLLM, Ollama, or hosted OpenAI-compatible
+ * endpoints.
+ */
+import type { ModelConfig } from './types.js';
+import type { ModelClient } from './executor.js';
+export interface OpenAICompatClientOptions {
+    /** Fallback endpoint when the ModelConfig has none (default: UAP_INFERENCE_ENDPOINT or http://localhost:4000/v1) */
+    defaultEndpoint?: string;
+    /** Request timeout in ms (default 300000) */
+    timeoutMs?: number;
+}
+export declare class OpenAICompatClient implements ModelClient {
+    private readonly defaultEndpoint;
+    private readonly timeoutMs;
+    constructor(options?: OpenAICompatClientOptions);
+    complete(model: ModelConfig, prompt: string, options?: {
+        maxTokens?: number;
+        timeout?: number;
+        temperature?: number;
+    }): Promise<{
+        content: string;
+        tokensUsed: {
+            input: number;
+            output: number;
+        };
+        latencyMs: number;
+    }>;
+}
+//# sourceMappingURL=openai-compat-client.d.ts.map

package/dist/models/openai-compat-client.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"openai-compat-client.d.ts","sourceRoot":"","sources":["../../src/models/openai-compat-client.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AAEH,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,YAAY,CAAC;AAC9C,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,eAAe,CAAC;AAEjD,MAAM,WAAW,yBAAyB;IACxC,oHAAoH;IACpH,eAAe,CAAC,EAAE,MAAM,CAAC;IACzB,6CAA6C;IAC7C,SAAS,CAAC,EAAE,MAAM,CAAC;CACpB;AAiBD,qBAAa,kBAAmB,YAAW,WAAW;IACpD,OAAO,CAAC,QAAQ,CAAC,eAAe,CAAS;IACzC,OAAO,CAAC,QAAQ,CAAC,SAAS,CAAS;gBAEvB,OAAO,GAAE,yBAA8B;IAQ7C,QAAQ,CACZ,KAAK,EAAE,WAAW,EAClB,MAAM,EAAE,MAAM,EACd,OAAO,CAAC,EAAE;QAAE,SAAS,CAAC,EAAE,MAAM,CAAC;QAAC,OAAO,CAAC,EAAE,MAAM,CAAC;QAAC,WAAW,CAAC,EAAE,MAAM,CAAA;KAAE,GACvE,OAAO,CAAC;QAAE,OAAO,EAAE,MAAM,CAAC;QAAC,UAAU,EAAE;YAAE,KAAK,EAAE,MAAM,CAAC;YAAC,MAAM,EAAE,MAAM,CAAA;SAAE,CAAC;QAAC,SAAS,EAAE,MAAM,CAAA;KAAE,CAAC;CA+DlG"}

package/dist/models/openai-compat-client.js

@@ -0,0 +1,82 @@
+/**
+ * OpenAI-compatible ModelClient
+ *
+ * Production implementation of the ModelClient interface from executor.ts.
+ * Talks to any /v1/chat/completions server — the local inference gateway
+ * (:4000), llama.cpp (:8080), vLLM, Ollama, or hosted OpenAI-compatible
+ * endpoints.
+ */
+const DEFAULT_TIMEOUT_MS = 300_000;
+const LOOPBACK_RE = /^(localhost|127\.\d+\.\d+\.\d+|\[?::1\]?)$/i;
+const PRIVATE_HOST_RE = /^(10\.|192\.168\.|172\.(1[6-9]|2\d|3[01])\.)/;
+function isLocalEndpoint(url) {
+    return LOOPBACK_RE.test(url.hostname) || PRIVATE_HOST_RE.test(url.hostname);
+}
+export class OpenAICompatClient {
+    defaultEndpoint;
+    timeoutMs;
+    constructor(options = {}) {
+        this.defaultEndpoint =
+            options.defaultEndpoint ??
+                process.env.UAP_INFERENCE_ENDPOINT ??
+                'http://localhost:4000/v1';
+        this.timeoutMs = options.timeoutMs ?? DEFAULT_TIMEOUT_MS;
+    }
+    async complete(model, prompt, options) {
+        const endpoint = (model.endpoint ?? this.defaultEndpoint).replace(/\/$/, '');
+        const apiKey = model.apiKeyEnvVar ? process.env[model.apiKeyEnvVar] : undefined;
+        const timeout = options?.timeout ?? this.timeoutMs;
+        const url = new URL(`${endpoint}/chat/completions`);
+        // Never send a credential in cleartext beyond the local network.
+        if (apiKey && url.protocol !== 'https:' && !isLocalEndpoint(url)) {
+            throw new Error(`Refusing to send ${model.apiKeyEnvVar} over ${url.protocol}// to non-local host ${url.hostname} — use https.`);
+        }
+        const start = Date.now();
+        const controller = new AbortController();
+        const timer = setTimeout(() => controller.abort(), timeout);
+        try {
+            const response = await fetch(url, {
+                method: 'POST',
+                headers: {
+                    'Content-Type': 'application/json',
+                    ...(apiKey ? { Authorization: `Bearer ${apiKey}` } : {}),
+                },
+                body: JSON.stringify({
+                    model: model.apiModel,
+                    messages: [{ role: 'user', content: prompt }],
+                    ...(options?.maxTokens ? { max_tokens: options.maxTokens } : {}),
+                    ...(options?.temperature !== undefined ? { temperature: options.temperature } : {}),
+                }),
+                signal: controller.signal,
+            });
+            if (!response.ok) {
+                const body = await response.text().catch(() => '');
+                throw new Error(`Inference request failed (${response.status}): ${body.slice(0, 500)}`);
+            }
+            const data = (await response.json());
+            const content = data.choices?.[0]?.message?.content;
+            if (typeof content !== 'string') {
+                const errMsg = typeof data.error === 'string' ? data.error : data.error?.message ?? 'no choices in response';
+                throw new Error(`Inference response missing content: ${errMsg}`);
+            }
+            return {
+                content,
+                tokensUsed: {
+                    input: data.usage?.prompt_tokens ?? 0,
+                    output: data.usage?.completion_tokens ?? 0,
+                },
+                latencyMs: Date.now() - start,
+            };
+        }
+        catch (err) {
+            if (err instanceof Error && err.name === 'AbortError') {
+                throw new Error(`Inference request timed out after ${timeout}ms (${url.hostname})`);
+            }
+            throw err;
+        }
+        finally {
+            clearTimeout(timer);
+        }
+    }
+}
+//# sourceMappingURL=openai-compat-client.js.map

package/dist/models/openai-compat-client.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"openai-compat-client.js","sourceRoot":"","sources":["../../src/models/openai-compat-client.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AAkBH,MAAM,kBAAkB,GAAG,OAAO,CAAC;AAEnC,MAAM,WAAW,GAAG,6CAA6C,CAAC;AAClE,MAAM,eAAe,GAAG,8CAA8C,CAAC;AAEvE,SAAS,eAAe,CAAC,GAAQ;IAC/B,OAAO,WAAW,CAAC,IAAI,CAAC,GAAG,CAAC,QAAQ,CAAC,IAAI,eAAe,CAAC,IAAI,CAAC,GAAG,CAAC,QAAQ,CAAC,CAAC;AAC9E,CAAC;AAED,MAAM,OAAO,kBAAkB;IACZ,eAAe,CAAS;IACxB,SAAS,CAAS;IAEnC,YAAY,UAAqC,EAAE;QACjD,IAAI,CAAC,eAAe;YAClB,OAAO,CAAC,eAAe;gBACvB,OAAO,CAAC,GAAG,CAAC,sBAAsB;gBAClC,0BAA0B,CAAC;QAC7B,IAAI,CAAC,SAAS,GAAG,OAAO,CAAC,SAAS,IAAI,kBAAkB,CAAC;IAC3D,CAAC;IAED,KAAK,CAAC,QAAQ,CACZ,KAAkB,EAClB,MAAc,EACd,OAAwE;QAExE,MAAM,QAAQ,GAAG,CAAC,KAAK,CAAC,QAAQ,IAAI,IAAI,CAAC,eAAe,CAAC,CAAC,OAAO,CAAC,KAAK,EAAE,EAAE,CAAC,CAAC;QAC7E,MAAM,MAAM,GAAG,KAAK,CAAC,YAAY,CAAC,CAAC,CAAC,OAAO,CAAC,GAAG,CAAC,KAAK,CAAC,YAAY,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;QAChF,MAAM,OAAO,GAAG,OAAO,EAAE,OAAO,IAAI,IAAI,CAAC,SAAS,CAAC;QAEnD,MAAM,GAAG,GAAG,IAAI,GAAG,CAAC,GAAG,QAAQ,mBAAmB,CAAC,CAAC;QACpD,iEAAiE;QACjE,IAAI,MAAM,IAAI,GAAG,CAAC,QAAQ,KAAK,QAAQ,IAAI,CAAC,eAAe,CAAC,GAAG,CAAC,EAAE,CAAC;YACjE,MAAM,IAAI,KAAK,CACb,oBAAoB,KAAK,CAAC,YAAY,SAAS,GAAG,CAAC,QAAQ,wBAAwB,GAAG,CAAC,QAAQ,eAAe,CAC/G,CAAC;QACJ,CAAC;QAED,MAAM,KAAK,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;QACzB,MAAM,UAAU,GAAG,IAAI,eAAe,EAAE,CAAC;QACzC,MAAM,KAAK,GAAG,UAAU,CAAC,GAAG,EAAE,CAAC,UAAU,CAAC,KAAK,EAAE,EAAE,OAAO,CAAC,CAAC;QAE5D,IAAI,CAAC;YACH,MAAM,QAAQ,GAAG,MAAM,KAAK,CAAC,GAAG,EAAE;gBAChC,MAAM,EAAE,MAAM;gBACd,OAAO,EAAE;oBACP,cAAc,EAAE,kBAAkB;oBAClC,GAAG,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,aAAa,EAAE,UAAU,MAAM,EAAE,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;iBACzD;gBACD,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC;oBACnB,KAAK,EAAE,KAAK,CAAC,QAAQ;oBACrB,QAAQ,EAAE,CAAC,EAAE,IAAI,EAAE,MAAM,EAAE,OAAO,EAAE,MAAM,EAAE,CAAC;oBAC7C,GAAG,CAAC,OAAO,EAAE,SAAS,CAAC,CAAC,CAAC,EAAE,UAAU,EAAE,OAAO,CAAC,SAAS,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;oBAChE,GAAG,CAAC,OAAO,EAAE,WAAW,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,WAAW,EAAE,OAAO,CAAC,WAAW,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;iBACpF,CAAC;gBACF,MAAM,EAAE,UAAU,CAAC,MAAM;aAC1B,CAAC,CAAC;YAEH,IAAI,CAAC,QAAQ,CAAC,EAAE,EAAE,CAAC;gBACjB,MAAM,IAAI,GAAG,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC,KAAK,CAAC,GAAG,EAAE,CAAC,EAAE,CAAC,CAAC;gBACnD,MAAM,IAAI,KAAK,CAAC,6BAA6B,QAAQ,CAAC,MAAM,MAAM,IAAI,CAAC,KAAK,CAAC,CAAC,EAAE,GAAG,CAAC,EAAE,CAAC,CAAC;YAC1F,CAAC;YAED,MAAM,IAAI,GAAG,CAAC,MAAM,QAAQ,CAAC,IAAI,EAAE,CAA2B,CAAC;YAC/D,MAAM,OAAO,GAAG,IAAI,CAAC,OAAO,EAAE,CAAC,CAAC,CAAC,EAAE,OAAO,EAAE,OAAO,CAAC;YACpD,IAAI,OAAO,OAAO,KAAK,QAAQ,EAAE,CAAC;gBAChC,MAAM,MAAM,GACV,OAAO,IAAI,CAAC,KAAK,KAAK,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,EAAE,OAAO,IAAI,wBAAwB,CAAC;gBAChG,MAAM,IAAI,KAAK,CAAC,uCAAuC,MAAM,EAAE,CAAC,CAAC;YACnE,CAAC;YAED,OAAO;gBACL,OAAO;gBACP,UAAU,EAAE;oBACV,KAAK,EAAE,IAAI,CAAC,KAAK,EAAE,aAAa,IAAI,CAAC;oBACrC,MAAM,EAAE,IAAI,CAAC,KAAK,EAAE,iBAAiB,IAAI,CAAC;iBAC3C;gBACD,SAAS,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,KAAK;aAC9B,CAAC;QACJ,CAAC;QAAC,OAAO,GAAG,EAAE,CAAC;YACb,IAAI,GAAG,YAAY,KAAK,IAAI,GAAG,CAAC,IAAI,KAAK,YAAY,EAAE,CAAC;gBACtD,MAAM,IAAI,KAAK,CAAC,qCAAqC,OAAO,OAAO,GAAG,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtF,CAAC;YACD,MAAM,GAAG,CAAC;QACZ,CAAC;gBAAS,CAAC;YACT,YAAY,CAAC,KAAK,CAAC,CAAC;QACtB,CAAC;IACH,CAAC;CACF"}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@miller-tech/uap",
-  "version": "1.26.6",
+  "version": "1.28.0",
   "description": "Autonomous AI agent memory system with CLAUDE.md protocol enforcement",
   "type": "module",
   "main": "dist/index.js",

package/tools/agents/docker-compose.qdrant.yml

@@ -2,13 +2,19 @@ version: '3.8'
 
 services:
   qdrant:
-    image: qdrant/qdrant:latest
+    # Pinned (was :latest, which drifts: a stale cached 1.16.3 broke client
+    # compatibility against @qdrant/js-client-rest 1.18.0, whose minor must be
+    # within 1 of the server). v1.18.1 matches the client minor.
+    image: qdrant/qdrant:v1.18.1
     container_name: uap-qdrant
     ports:
       - '6333:6333' # REST API
       - '6334:6334' # gRPC
     volumes:
       - ./qdrant_storage:/qdrant/storage:z
+      # Persist snapshots on the host (default /qdrant/snapshots is ephemeral,
+      # so backups are lost on container recreate).
+      - ./qdrant_snapshots:/qdrant/snapshots:z
     environment:
       - QDRANT__SERVICE__GRPC_PORT=6334
       - QDRANT__REST__PORT=6333