From 7070342fb1d0852be513e9ef4457404033a86a6f Mon Sep 17 00:00:00 2001
From: robbond <rob@rdbsolutions.co.uk>
Date: Sat, 5 Sep 2026 19:41:01 +0100
Subject: [PATCH] fix(confidence-engine): preserve successful chat detection

---
 docs/current-handoff.md    | 6 ++++++
 lib/llm/provider.js        | 8 ++++----
 tests/llm/provider.test.js | 8 ++++----
 3 files changed, 14 insertions(+), 8 deletions(-)

diff --git a/docs/current-handoff.md b/docs/current-handoff.md
index 2e7479b..cccea91 100644
--- a/docs/current-handoff.md
+++ b/docs/current-handoff.md
@@ -92,6 +92,12 @@
 - Provider execution diagnostics now distinguish chat skipped due capability state, chat attempted and failed before generate fallback, and successful chat without fallback. They are deterministically verified through `/api/cases/start`; endpoint selection and fallback behaviour remain unchanged.
 - Next restart point: one observation-only fixed-scenario production call to identify why `/api/generate` is reached.
 
+## Chat capability body disposal
+
+- A fresh-process debugger observation proved capability `/api/chat` returned HTTP 200, then invalid `res.body?.consume()` threw and the catch incorrectly cached `_chatSupported = false`.
+- This skipped schema-constrained reconstruction `/api/chat` and used unconstrained `/api/generate`. Response-body disposal now uses Fetch-compatible consumption without changing capability or fallback semantics.
+- Next boundary: one fresh-process fixed-scenario production observation.
+
 ## Current product architecture
 
 Three distinct routes, not a single page:
diff --git a/lib/llm/provider.js b/lib/llm/provider.js
index fcca1f3..4bdb74e 100644
--- a/lib/llm/provider.js
+++ b/lib/llm/provider.js
@@ -84,13 +84,13 @@ async function detectChatSupport(baseUrl, modelName) {
     });
     
     if (res.ok) {
-      await res.body?.consume();
+      await res.text();
       _chatSupported = true;
     } else if (res.status === 405 || res.status === 501) {
-      await res.body?.consume();
+      await res.text();
       _chatSupported = false;
     } else {
-      await res.body?.consume();
+      await res.text();
       _chatSupported = false;
     }
   } catch {
@@ -160,7 +160,7 @@ class OllamaLlmProvider {
             : JSON.stringify(fullResponseData.message?.content ?? null);
           apiUsed = "/api/chat";
         } else {
-          await res.body?.consume();
+          await res.text();
         }
       } catch (e) {
         if (!e.message.includes("abort")) { /* non-fatal */ }
diff --git a/tests/llm/provider.test.js b/tests/llm/provider.test.js
index e23a198..eb16177 100644
--- a/tests/llm/provider.test.js
+++ b/tests/llm/provider.test.js
@@ -5,7 +5,7 @@ describe("OllamaLlmProvider chat capability detection", () => {
   it("uses the configured model for the chat probe and keeps the chat path", async () => {
     const originalBaseUrl = process.env.OLLAMA_BASE_URL;
     const fetchSpy = vi.fn()
-      .mockResolvedValueOnce({ ok: true, body: { consume: vi.fn() } })
+      .mockResolvedValueOnce({ ok: true, text: async () => "" })
       .mockResolvedValueOnce({
         ok: true,
         json: async () => ({ message: { content: "{}" } }),
@@ -55,7 +55,7 @@ describe("OllamaLlmProvider chat capability detection", () => {
   it("reports chat-skipped generate fallback execution", async () => {
     const originalBaseUrl = process.env.OLLAMA_BASE_URL;
     const fetchSpy = vi.fn()
-      .mockResolvedValueOnce({ ok: false, status: 501, body: { consume: vi.fn() } })
+      .mockResolvedValueOnce({ ok: false, status: 501, text: async () => "" })
       .mockResolvedValueOnce({ ok: false, status: 500, text: async () => "failure" });
     vi.stubGlobal("fetch", fetchSpy);
     process.env.OLLAMA_BASE_URL = "http://ollama.test";
@@ -84,8 +84,8 @@ describe("OllamaLlmProvider chat capability detection", () => {
   it("reports chat-attempt-failed generate fallback execution", async () => {
     const originalBaseUrl = process.env.OLLAMA_BASE_URL;
     const fetchSpy = vi.fn()
-      .mockResolvedValueOnce({ ok: true, body: { consume: vi.fn() } })
-      .mockResolvedValueOnce({ ok: false, body: { consume: vi.fn() } })
+      .mockResolvedValueOnce({ ok: true, text: async () => "" })
+      .mockResolvedValueOnce({ ok: false, text: async () => "" })
       .mockResolvedValueOnce({ ok: false, status: 500, text: async () => "failure" });
     vi.stubGlobal("fetch", fetchSpy);
     process.env.OLLAMA_BASE_URL = "http://ollama.test";