Make Ollama context size and threads configurable

tooling-container/VERSION

@@ -1 +1 @@
-0.1.0
+0.1.1

tooling-container/src/extract.py

@@ -10,6 +10,11 @@ import requests
 
 OLLAMA_URL = os.environ.get("OLLAMA_URL", "http://127.0.0.1:11434/api/chat")
 MODEL = os.environ.get("OLLAMA_MODEL", "qwen2.5:3b")
+# Ollama restarts its runner when a request's context size or thread count
+# differs from the loaded one; set these to match other clients of a shared
+# Ollama. OLLAMA_NUM_THREAD unset = Ollama's default.
+NUM_CTX = int(os.environ.get("OLLAMA_NUM_CTX", "3072"))
+NUM_THREAD = os.environ.get("OLLAMA_NUM_THREAD")
 
 SCHEMA = {
     "type": "object",
@@ -57,6 +62,9 @@ Respond with ONLY the JSON object, no other text.
 
 
 def structure(ocr_text: str) -> dict:
+    options = {"temperature": 0, "num_ctx": NUM_CTX}
+    if NUM_THREAD:
+        options["num_thread"] = int(NUM_THREAD)
     payload = {
         "model": MODEL,
         "messages": [
@@ -64,7 +72,7 @@ def structure(ocr_text: str) -> dict:
         ],
         "format": SCHEMA,
         "stream": False,
-        "options": {"temperature": 0, "num_ctx": 3072},
+        "options": options,
     }
 
     start = time.time()