@canarycoders/ai 0.6.0 → 0.8.0

package/README.md

@@ -4,7 +4,7 @@ TypeScript SDK for the [CanaryCoders AI](https://ai.canarycoders.es) gateway. Ru
 
 **Docs:** https://docs.ai.canarycoders.es
 
-One client covers chat, image/video/audio generation, embeddings, vision, conversational agents, realtime sessions, and usage data. The queue polling, SSE streaming, retries, and error typing are handled for you, so most calls are a single `await`.
+One client covers chat, image/video/audio generation, embeddings, vision, OCR/PII-redaction, conversational agents, realtime sessions, and usage data. The queue polling, SSE streaming, retries, and error typing are handled for you, so most calls are a single `await`.
 
 ## Install
 
@@ -77,6 +77,20 @@ const song = await client.audio.music({ prompt: "calm lo-fi", durationMs: 20000
 const { detections } = await client.vision.detect({ image });
 ```
 
+## OCR and PII redaction (local)
+
+Text recognition with bounding boxes and PII redaction, processed fully on CanaryLLM's own EU infrastructure (Apple Vision). Images and extracted text are never stored.
+
+```ts
+const { text, lines } = await client.ocr.recognize({ image, granularity: "word" });
+
+const { redactedImage, counts } = await client.ocr.redact({
+  image,
+  presets: ["email", "phone", "iban"],
+  style: "blur", // "box" is the safest for hard PII redaction
+});
+```
+
 ## Embeddings
 
 Embed text into vectors with a local model (LM Studio), for customer-side RAG. The gateway processes the text transiently and stores nothing — you keep the vectors and documents in your own store (e.g. pgvector, sqlite-vec).

package/dist/index.cjs

@@ -682,6 +682,7 @@ async function* nativeQueueAdapter(frames, includeRaw) {
   const raw = (v) => includeRaw ? v : void 0;
   let lastUsage;
   let lastFinish;
+  let lastCost;
   for await (const frame of frames) {
     switch (frame.event) {
       case "start":
@@ -690,11 +691,12 @@ async function* nativeQueueAdapter(frames, includeRaw) {
       case "error":
         throw streamError(tryParse(frame.data));
       case "done":
-        yield { type: "done", finishReason: lastFinish, usage: lastUsage };
+        yield { type: "done", finishReason: lastFinish, usage: lastUsage, cost: lastCost };
         return;
       case "chunk": {
         const chunk = tryParse(frame.data);
         if (!chunk) break;
+        if (chunk.cost) lastCost = chunk.cost;
         if (chunk.usage) {
           lastUsage = chunk.usage;
           yield { type: "usage", usage: chunk.usage, raw: raw(chunk) };
@@ -722,7 +724,7 @@ async function* nativeQueueAdapter(frames, includeRaw) {
       }
     }
   }
-  yield { type: "done", finishReason: lastFinish, usage: lastUsage };
+  yield { type: "done", finishReason: lastFinish, usage: lastUsage, cost: lastCost };
 }
 async function* openAIAdapter(frames, includeRaw) {
   const raw = (v) => includeRaw ? v : void 0;
@@ -1588,6 +1590,27 @@ var VisionResource = class extends BaseResource {
   }
 };
 
+// src/resources/ocr.ts
+var OcrResource = class extends BaseResource {
+  recognize(params, poll) {
+    return this.runQueued("/api/ocr/recognize", params, "vision", poll);
+  }
+  recognizeJob(params, signal) {
+    return this.submitQueued("/api/ocr/recognize", params, "vision", signal);
+  }
+  redact(params, poll) {
+    return this.runQueued("/api/ocr/redact", params, "vision", poll);
+  }
+  redactJob(params, signal) {
+    return this.submitQueued("/api/ocr/redact", params, "vision", signal);
+  }
+  async languages(signal) {
+    return this.transport.json("GET", "/api/ocr/languages", {
+      signal
+    });
+  }
+};
+
 // src/client.ts
 var DEFAULT_BASE_URL = "https://api.ai.canarycoders.es";
 function readEnv(key) {
@@ -1602,6 +1625,7 @@ var CanaryCodersAI = class {
   audio;
   embeddings;
   vision;
+  ocr;
   conversations;
   agents;
   realtime;
@@ -1634,6 +1658,7 @@ var CanaryCodersAI = class {
     this.audio = new AudioResource(this.transport, poll);
     this.embeddings = new EmbeddingsResource(this.transport, poll);
     this.vision = new VisionResource(this.transport, poll);
+    this.ocr = new OcrResource(this.transport, poll);
     this.conversations = new ConversationsResource(this.transport, poll);
     this.agents = new AgentsResource(this.transport, poll);
     this.realtime = new RealtimeResource(this.transport, poll);