@orkestrel/ollama 0.0.13 → 0.0.15

package/README.md

@@ -1,9 +1,15 @@
 # @orkestrel/ollama
 
-A typed local-LLM provider for the `@orkestrel` line — a `ProviderInterface`
-implementation over a local Ollama daemon's `POST /api/chat`, with NDJSON
-streaming, tool calls, thinking, and usage accounting, built on pure
-web-standard `fetch` / `ReadableStream` (no Ollama SDK dependency).
+> A typed local-LLM provider for the `@orkestrel` line: a `ProviderInterface` over a
+> local Ollama daemon's `POST /api/chat`, with non-streaming `generate`, NDJSON
+> `stream`, tool calls, thinking, and usage accounting, built on web-standard `fetch`
+> and `ReadableStream` with no Ollama SDK dependency.
+
+Create a provider with the `createOllama` function, hand it a conversation and a
+bounding `AbortSignal`, and read the assembled `ProviderResult` the `generate`
+method resolves — or drive the `stream` method for live deltas. Point `url` at your
+own server and attach a short-lived token through `headers` where a browser runtime
+must not hold the real key.
 
 ## Install
 
@@ -15,8 +21,9 @@ npm install @orkestrel/ollama
 
 - Node.js >= 22
 - A running Ollama daemon (default `http://localhost:11434`) with a pulled
-  model — required at runtime by any consumer, and by this repo's live-only
-  `src:server` test suite
+  model — required at runtime by any consumer, and by this repository's live
+  `service` test project; the `src:server` project is hermetic and passes with
+  the daemon down
 - ESM + CJS (dual-format build)
 
 ## Usage
@@ -30,7 +37,7 @@ const abort = createAbort()
 const messages = [{ id: '1', role: 'user', content: 'Reply with exactly: ok' }] as const
 
 const result = await provider.generate(messages, abort.signal)
-result.content // 'ok'
+result.content // the assistant's answer text
 result.usage // { prompt, completion, total }
 ```
 
@@ -43,7 +50,7 @@ assembled result when the stream completes:
 const generator = provider.stream(messages, abort.signal)
 let step = await generator.next()
 while (!step.done) {
-	if (step.value.type === 'content') process.stdout.write(step.value.text)
+	if (step.value.channel === 'content') process.stdout.write(step.value.text)
 	step = await generator.next()
 }
 const streamed = step.value // the assembled ProviderResult
@@ -54,7 +61,7 @@ const streamed = step.value // the assembled ProviderResult
 For the full surface — `createOllama`, `OllamaProvider`, `OllamaOptions`,
 tool calls (`ToolDefinition` / `ToolCall` from `@orkestrel/tool`), thinking,
 and the context-framing default — see
-[`guides/src/ollama.md`](guides/src/ollama.md).
+[`guides/ollama.md`](guides/ollama.md).
 
 ## Package