@pairsystems/goodmem 0.1.0

package/README.md

@@ -0,0 +1,79 @@
+# GoodMem TypeScript SDK
+
+Generated v2 TypeScript SDK for the GoodMem API.
+
+```ts
+import { Goodmem } from "@pairsystems/goodmem";
+
+const client = new Goodmem({
+  baseUrl: "http://localhost:8080",
+  apiKey: process.env.GOODMEM_API_KEY,
+});
+
+const embedder = await client.embedders.create(
+  {
+    displayName: "OpenAI Small",
+    modelIdentifier: "text-embedding-3-small",
+  },
+  { apiKey: process.env.OPENAI_API_KEY },
+);
+
+const spaces = await client.spaces.list();
+
+await client.memories.create({
+  spaceId: "space-id",
+  originalContent: "hello", // contentType defaults to text/plain.
+});
+
+await client.memories.createFromFile({
+  spaceId: "space-id",
+  file: new Blob(["hello"], { type: "text/plain" }),
+  filename: "note.txt",
+});
+
+await client.memories.createFromBytes({
+  spaceId: "space-id",
+  bytes: new Uint8Array([1, 2, 3]),
+  contentType: "application/octet-stream",
+  filename: "payload.bin",
+});
+
+await client.memories.createFromPath({
+  spaceId: "space-id",
+  path: "./notes.txt", // contentType inferred as text/plain; explicit contentType overrides.
+});
+
+const ocr = await client.ocr.documentFromFile({
+  file: new Blob([pdfBytes], { type: "application/pdf" }),
+  format: "PDF",
+  includeMarkdown: true,
+});
+```
+
+The provider, endpoint, dimensionality, distribution type, and supported modalities
+for known embedders, LLMs, and rerankers are inferred from the bundled model
+registry. Pass explicit fields to override registry defaults or to use a custom
+model.
+
+For file ingestion, `createFromFile` sends multipart form data and infers
+`contentType` from `Blob.type`/`File.type` when available. `createFromBytes`
+accepts `Uint8Array` or `ArrayBuffer` and requires an explicit `contentType`.
+In Node.js, `createFromPath` uses `fs.openAsBlob` when available, falls back to
+`fs.promises.readFile`, and infers `contentType` from common file extensions.
+Unknown or extensionless paths default to `application/octet-stream`; pass
+`contentType` explicitly to override inference.
+
+JSON model fields that represent raw API `bytes`, such as
+`Memory.originalContent`, are base64-encoded strings on the JSON wire.
+
+For OCR, `documentFromFile`, `documentFromBytes`, and Node-only
+`documentFromPath` base64-encode bytes into the existing JSON OCR endpoint; OCR
+does not use multipart upload.
+
+Runtime support: Node.js 20+ is supported. Browser and worker runtimes must
+provide `fetch`, `Headers`, `Response`, `ReadableStream`, `Blob`, `FormData`, and
+`TextEncoder`. `memories.createFromPath` and `ocr.documentFromPath` are Node-only
+because they use `node:fs`.
+
+This package is generated from `clients/v2/python/goodmem/api_ir.json`.
+Do not edit generated files under `src/` manually; run `../clients_gen.sh`.