From 4479683cf44980df8b7d670b85c87544138ffa6d Mon Sep 17 00:00:00 2001
From: mptyl <mp@tylconsulting.it>
Date: Thu, 2 Jul 2026 11:49:11 +0200
Subject: [PATCH] chore(pi): scope AritmoLab provider to project, drop Gemma
 model

Move the AritmoLab provider (Qwen) into harness/.pi/extensions/ so it is
auto-discovered only when Pi runs with cwd=harness, keeping it project-local.
GLM (~/.pi/agent/models.json) and DeepSeek (built-in) stay user-wide. File is
.js (not .mjs) because Pi extension auto-discovery matches only /\.(ts|js)$/.

Removes the gemma4-26b-a4b model (404 at the endpoint) from both the provider
and the model-matrix default list.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---
 harness/.pi/extensions/aritmolab-provider.js | 96 ++++++++++++++++++++
 harness/scripts/model-matrix.mjs             |  2 +-
 2 files changed, 97 insertions(+), 1 deletion(-)
 create mode 100644 harness/.pi/extensions/aritmolab-provider.js

diff --git a/harness/.pi/extensions/aritmolab-provider.js b/harness/.pi/extensions/aritmolab-provider.js
new file mode 100644
index 00000000..f9f43f08
--- /dev/null
+++ b/harness/.pi/extensions/aritmolab-provider.js
@@ -0,0 +1,96 @@
+import { createAssistantMessageEventStream, streamSimpleOpenAICompletions } from "@mariozechner/pi-ai";
+
+function convertThinkingBlocks(message) {
+  return {
+    ...message,
+    content: (message.content ?? []).map((block) => {
+      if (block?.type !== "thinking") return block;
+      return { type: "text", text: block.thinking ?? "" };
+    }),
+  };
+}
+
+function convertEvent(event) {
+  if (event.type === "thinking_start") {
+    return { type: "text_start", contentIndex: event.contentIndex, partial: convertThinkingBlocks(event.partial) };
+  }
+  if (event.type === "thinking_delta") {
+    return { type: "text_delta", contentIndex: event.contentIndex, delta: event.delta, partial: convertThinkingBlocks(event.partial) };
+  }
+  if (event.type === "thinking_end") {
+    return { type: "text_end", contentIndex: event.contentIndex, content: event.content, partial: convertThinkingBlocks(event.partial) };
+  }
+  if (event.type === "done") {
+    return { ...event, message: convertThinkingBlocks(event.message) };
+  }
+  if (event.type === "error") {
+    return { ...event, error: convertThinkingBlocks(event.error) };
+  }
+  if (event.partial) {
+    return { ...event, partial: convertThinkingBlocks(event.partial) };
+  }
+  return event;
+}
+
+function streamAritmolab(model, context, options) {
+  const source = streamSimpleOpenAICompletions(model, context, options);
+  const stream = createAssistantMessageEventStream();
+
+  (async () => {
+    try {
+      for await (const event of source) {
+        stream.push(convertEvent(event));
+      }
+    } catch (error) {
+      const message = {
+        role: "assistant",
+        content: [],
+        api: model.api,
+        provider: model.provider,
+        model: model.id,
+        usage: {
+          input: 0,
+          output: 0,
+          cacheRead: 0,
+          cacheWrite: 0,
+          totalTokens: 0,
+          cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },
+        },
+        stopReason: options?.signal?.aborted ? "aborted" : "error",
+        errorMessage: error instanceof Error ? error.message : String(error),
+        timestamp: Date.now(),
+      };
+      stream.push({ type: "error", reason: message.stopReason, error: message });
+      stream.end();
+    }
+  })();
+
+  return stream;
+}
+
+export default function (pi) {
+  pi.registerProvider("aritmolab", {
+    name: "AritmoLab",
+    baseUrl: "https://ml-aritmolab.policlinicosandonato.it/v1",
+    apiKey: "aritmolab",
+    api: "openai-completions",
+    streamSimple: streamAritmolab,
+    models: [
+      {
+        id: "qwen3.6-35b-a3b",
+        name: "AritmoLab Qwen3.6 35B A3B",
+        reasoning: false,
+        input: ["text"],
+        cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
+        contextWindow: 131072,
+        maxTokens: 16384,
+        compat: {
+          supportsDeveloperRole: false,
+          supportsReasoningEffort: false,
+          supportsStore: false,
+          maxTokensField: "max_tokens",
+        },
+      },
+    ],
+  });
+}
diff --git a/harness/scripts/model-matrix.mjs b/harness/scripts/model-matrix.mjs
index f0f7c0f1..f555ce61 100644
--- a/harness/scripts/model-matrix.mjs
+++ b/harness/scripts/model-matrix.mjs
@@ -32,7 +32,7 @@ if (!sessionId) {
   process.exit(2);
 }
 const models = (process.argv[3] ??
-  "zai/glm-5.2,deepseek/deepseek-v4-pro,aritmolab/qwen3.6-35b-a3b,aritmolab/gemma4-26b-a4b")
+  "zai/glm-5.2,deepseek/deepseek-v4-pro,aritmolab/qwen3.6-35b-a3b")
   .split(",").map((s) => s.trim()).filter(Boolean);
 const thinking = process.argv[4] ?? "medium";
 const outDir = process.argv[5] ?? "/tmp/model-matrix";