internal/adapters/llm/client.go

@@ -55,6 +55,29 @@ func WithMaxTokens(n int) Option {
 	}
 }
 
+// maxErrorBody bounds how much of a failing response ends up in an error message.
+const maxErrorBody = 512
+
+// HTTPError is a non-200 answer from the gateway. Its text keeps the form callers that
+// log it already rely on, with the body bounded; its Code is what code may branch on and
+// what errkind turns into a kind. The body is provider text and can echo the prompt, so
+// nothing private may persist or display Error() (ADR-035 review F3).
+type HTTPError struct {
+	Code int
+	Body string
+}
+
+func (e *HTTPError) Error() string {
+	body := e.Body
+	if len(body) > maxErrorBody {
+		body = body[:maxErrorBody] + "…"
+	}
+	return fmt.Sprintf("LLM returned %d: %s", e.Code, body)
+}
+
+// HTTPStatus implements errkind.HTTPStatuser.
+func (e *HTTPError) HTTPStatus() int { return e.Code }
+
 // WithTemperature overrides the sampling temperature. Zero IS a valid value and is sent
 // as zero: the reconciler's accuracy was measured at temperature 0 and must run there.
 func WithTemperature(t float64) Option {
@@ -166,7 +189,7 @@ func (c *Client) Complete(ctx context.Context, system, user string) (string, err
 		return "", fmt.Errorf("read response: %w", err)
 	}
 	if resp.StatusCode != http.StatusOK {
-		return "", fmt.Errorf("LLM returned %d: %s", resp.StatusCode, out)
+		return "", &HTTPError{Code: resp.StatusCode, Body: string(out)}
 	}
 
 	var cr chatResponse