agent-lattice 0.17.0 → 0.18.0

package/README.md

@@ -140,10 +140,16 @@ const result = await agent.prompt("Solve this carefully.", {
 
 Use `{ type: "adaptive" }` for models that support adaptive thinking. Use
 `{ type: "enabled", budgetTokens }` for models that require a fixed budget, or
-`{ type: "disabled" }` to omit thinking from the provider request. A fixed
+`{ type: "disabled" }` to turn thinking off. A fixed
 budget is capped at `maxTokens - 1` to satisfy the Anthropic API constraint.
 When omitted, the SDK does not send a thinking configuration.
 
+`{ type: "disabled" }` is sent to the provider explicitly as
+`thinking: { "type": "disabled" }` rather than omitted, because some
+Anthropic-compatible providers (for example DeepSeek's
+`https://api.deepseek.com/anthropic` endpoint) default thinking to on —
+omitting the field would leave it enabled.
+
 For Kimi K3 through an Anthropic-compatible endpoint or gateway, use
 `reasoningEffort` to send the provider's top-level `reasoning_effort` parameter:
 
@@ -1295,6 +1301,13 @@ if (result.stop_reason === "max_tokens") {
 The SDK does not treat that as an error, so checking this field is the only way
 to distinguish a complete answer from a truncated one.
 
+When a response containing tool calls is truncated at `max_tokens`, the SDK
+does not execute those calls: the last `tool_use` input may be incomplete, and
+a truncated value can even survive JSON parsing with its meaning changed.
+Every call in the batch gets an error `tool_result` explaining the truncation
+and asking the model to reissue the call with a shorter output, and the loop
+continues. *Requires 0.18.0 or later.*
+
 Usage comes from the model client. The built-in Anthropic client fills it in from
 the response, including the streaming path; a custom `ModelClient` that omits
 `usage` produces zeroed counts rather than an error.