iints-sdk-python35 1.1.2__py3-none-any.whl → 1.1.3__py3-none-any.whl

iints/__init__.py

@@ -11,7 +11,7 @@ except ImportError:  # pragma: no cover - Python < 3.8 fallback
 try:
     __version__ = version("iints-sdk-python35")
 except PackageNotFoundError:  # pragma: no cover - source tree fallback
-    __version__ = "1.1.2"
+    __version__ = "1.1.3"
 
 # Note to developers: this SDK is currently maintained by a single author.
 # Please report bugs via GitHub issues and feel free to contribute fixes via PRs.

iints/ai/backends/ollama.py

@@ -2,6 +2,8 @@ from __future__ import annotations
 
 import json
 import os
+from http.client import IncompleteRead, RemoteDisconnected
+from time import sleep
 from urllib import error, request
 
 
@@ -40,6 +42,20 @@ class OllamaBackend:
     def _pull_hint(self) -> str:
         return f"ollama pull {self.model_name}"
 
+    def _generation_failure_hint(self) -> str:
+        resolved = self.resolved_model_name or self.model_name
+        return (
+            "Ollama closed the generation connection before returning a response.\n"
+            f"Endpoint: {self.base_url}\n"
+            f"Model: {resolved}\n"
+            "This usually means the model crashed while loading, the daemon restarted, "
+            "or the machine ran out of memory.\n"
+            "Try one of these:\n"
+            f"  1. Run `ollama run {resolved} \"Reply with OK.\"` to confirm direct inference works.\n"
+            "  2. Run `iints ai local-check --smoke-test` to validate a real generation path.\n"
+            "  3. Switch to a smaller local model such as `ministral-3:3b` if memory is tight."
+        )
+
     def _requires_ministral_3_runtime(self) -> bool:
         requested = self.model_name.strip().lower()
         return requested.startswith("ministral-3") or requested == "ministral"
@@ -68,6 +84,15 @@ class OllamaBackend:
         payload: dict[str, object] | None = None,
         *,
         method: str = "POST",
+    ) -> dict[str, object]:
+        return self._request_json_once(path, payload, method=method)
+
+    def _request_json_once(
+        self,
+        path: str,
+        payload: dict[str, object] | None = None,
+        *,
+        method: str = "POST",
     ) -> dict[str, object]:
         url = f"{self.base_url}{path}"
         body = None
@@ -92,6 +117,12 @@ class OllamaBackend:
                 f"Could not reach Ollama at {self.base_url}. "
                 "Start Ollama or set OLLAMA_HOST to the correct endpoint."
             ) from exc
+        except (RemoteDisconnected, ConnectionResetError, IncompleteRead) as exc:
+            if path == "/api/generate":
+                raise RuntimeError(self._generation_failure_hint()) from exc
+            raise RuntimeError(
+                f"Ollama connection closed unexpectedly while calling {path} at {self.base_url}."
+            ) from exc
 
         try:
             payload_json = json.loads(text)
@@ -223,6 +254,43 @@ class OllamaBackend:
             "version_ok": version_ok,
         }
 
+    def smoke_test(self) -> dict[str, object]:
+        resolved_model = self.ensure_model_ready()
+        payload = {
+            "model": resolved_model,
+            "system": "You are a health check. Reply with exactly: OK",
+            "prompt": "Reply with exactly: OK",
+            "stream": False,
+            "options": {
+                "temperature": 0,
+                "num_predict": 8,
+            },
+        }
+
+        last_error: Exception | None = None
+        for attempt in range(2):
+            try:
+                response = self._request_json_once("/api/generate", payload)
+                text = response.get("response")
+                if not isinstance(text, str) or not text.strip():
+                    raise RuntimeError("Ollama returned an empty smoke-test completion.")
+                return {
+                    "ok": True,
+                    "response": text.strip(),
+                    "attempts": attempt + 1,
+                }
+            except (RuntimeError, RemoteDisconnected, ConnectionResetError, IncompleteRead) as exc:
+                if not isinstance(exc, RuntimeError):
+                    exc = RuntimeError(self._generation_failure_hint())
+                last_error = exc
+                if attempt == 0:
+                    sleep(1.0)
+                    continue
+                break
+
+        assert last_error is not None
+        raise last_error
+
     def complete(self, *, system_prompt: str, user_prompt: str) -> str:
         resolved_model = self.ensure_model_ready()
         payload = {
@@ -231,7 +299,22 @@ class OllamaBackend:
             "prompt": user_prompt,
             "stream": False,
         }
-        response = self._request_json("/api/generate", payload)
+        last_error: Exception | None = None
+        for attempt in range(2):
+            try:
+                response = self._request_json_once("/api/generate", payload)
+                break
+            except (RuntimeError, RemoteDisconnected, ConnectionResetError, IncompleteRead) as exc:
+                if not isinstance(exc, RuntimeError):
+                    exc = RuntimeError(self._generation_failure_hint())
+                last_error = exc
+                if attempt == 0:
+                    sleep(1.0)
+                    continue
+                raise exc
+        else:
+            assert last_error is not None
+            raise last_error
         text = response.get("response")
         if not isinstance(text, str) or not text.strip():
             raise RuntimeError("Ollama returned an empty completion.")

iints/ai/cli.py

@@ -111,6 +111,7 @@ def _render_local_check(console: Console, status: dict[str, object]) -> None:
     installed_text = ", ".join(str(item) for item in installed) if isinstance(installed, list) and installed else "none"
     ready = bool(status.get("ready"))
     resolved_model = status.get("resolved_model") or "not found"
+    smoke_text = status.get("smoke_test") or "not run"
     console.print(
         Panel(
             "\n".join(
@@ -126,6 +127,7 @@ def _render_local_check(console: Console, status: dict[str, object]) -> None:
                         if status.get("pull_command")
                         else "Pull command: not needed"
                     ),
+                    f"Generate smoke-test: {smoke_text}",
                 ]
             ),
             title="IINTS AI Local Check",
@@ -239,6 +241,13 @@ def local_check(
     model: Annotated[str, typer.Option(help="Ollama model name to validate locally.")] = DEFAULT_MINISTRAL_MODEL,
     ollama_host: Annotated[Optional[str], typer.Option(help="Override the Ollama base URL.")] = None,
     timeout_seconds: Annotated[float, typer.Option(help="HTTP timeout for Ollama health checks.")] = 120.0,
+    smoke_test: Annotated[
+        bool,
+        typer.Option(
+            "--smoke-test/--no-smoke-test",
+            help="Run a tiny generation request after health checks to prove the model can actually answer.",
+        ),
+    ] = True,
 ) -> None:
     console = Console()
     backend = OllamaBackend(model_name=model, base_url=ollama_host, timeout_seconds=timeout_seconds)
@@ -250,6 +259,13 @@ def local_check(
             )
             raise typer.Exit(code=1)
         status = backend.healthcheck()
+        if smoke_test and bool(status.get("ready")):
+            smoke = backend.smoke_test()
+            status["smoke_test"] = f"OK ({smoke.get('response')})"
+        elif smoke_test:
+            status["smoke_test"] = "skipped (model not ready)"
+        else:
+            status["smoke_test"] = "disabled"
         _render_local_check(console, status)
         if not bool(status.get("ready")):
             raise typer.Exit(code=1)

{iints_sdk_python35-1.1.2.dist-info → iints_sdk_python35-1.1.3.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: iints-sdk-python35
-Version: 1.1.2
+Version: 1.1.3
 Summary: A pre-clinical Edge-AI SDK for diabetes management validation.
 Author-email: Rune Bobbaers <rune.bobbaers@gmail.com>
 Project-URL: Homepage, https://github.com/python35/IINTS-SDK
@@ -110,6 +110,8 @@ ollama pull ministral-3:8b
 iints ai local-check --model ministral-3:8b
 ```
 
+`local-check` now performs a tiny generation smoke-test by default, so it verifies both model presence and real inference readiness.
+
 Recommended flow:
 
 ```bash
@@ -134,6 +136,7 @@ Notes:
 - Users can choose a larger or smaller local Mistral-family model with `--model ...`.
 - Large JSON payloads are clipped automatically before prompt generation to keep local inference stable.
 - `iints ai prepare <run_dir>` now creates AI-ready JSON payloads and, when MDMP is installed, a local development certificate plus keypair in `<run_dir>/ai/`.
+- If Ollama closes the connection during generation, the SDK now surfaces an explicit recovery hint and points users toward `ministral-3:3b` for lower-memory systems.
 - After `iints ai prepare`, you can point `iints ai explain|trends|anomalies|report` directly at the run directory.
 - Output is research-only and not medical advice.
 
@@ -144,7 +147,7 @@ Troubleshooting:
 
 ```bash
 python -m pip uninstall -y iints iints-sdk-python35
-python -m pip install -U "iints-sdk-python35[mdmp]==1.1.2"
+python -m pip install -U "iints-sdk-python35[mdmp]==1.1.3"
 hash -r
 ```
 

{iints_sdk_python35-1.1.2.dist-info → iints_sdk_python35-1.1.3.dist-info}/RECORD

@@ -1,9 +1,9 @@
-iints/__init__.py,sha256=wfAcfS7htgnV4JD-R8_WyKZHOwR8Z98vy-oLYxu0-rE,6391
+iints/__init__.py,sha256=F_rnQgeZZn3bQpyYAy4JwHARfQfq8fUSLvbs07Ycybw,6391
 iints/highlevel.py,sha256=DX12LRmL6YaYY99P0c_P93xfHe4mZjqyLhTYuS6L6hI,20491
 iints/metrics.py,sha256=O9hqOqJpUhUJDqsbfuqRMS9dkV97gzcgh3Y2jYUqHzg,907
 iints/ai/__init__.py,sha256=nyRDcFfSHI4a3NbTvySipFc3_inqRMEsr6xIEipWuyo,575
 iints/ai/assistant.py,sha256=0Ye1IaWEYg2rZnk3ny8f0GMoYqOWIa7U_GsV-sWrxtU,4346
-iints/ai/cli.py,sha256=_1ogEAb36BAt7sZ2CQSRKIJSpdn5xrlM7nNTtBIqfRo,18345
+iints/ai/cli.py,sha256=jR9cgLkYGQRx_AfiGGWK4oBFTQDO2fHas_3UJGq78X8,19009
 iints/ai/mdmp_guard.py,sha256=BpFQX0oyP9WMCUZbFhhoBzomNeVKuI1HY1EFH9cG8EE,4249
 iints/ai/model_catalog.py,sha256=gRW-i4eaXkrjX3mIKJlGzHqzU75lpIulEFKQsCX11CI,1804
 iints/ai/prepare.py,sha256=z3y5elCAMv0p_aNq4gQfZA1uIT7_cX3FGRdzmoZoKho,12967
@@ -11,7 +11,7 @@ iints/ai/prompts.py,sha256=pGp9tC1wBZXGG5duxfktaJEF4p_cvmR0zEIxmMTEAyE,2812
 iints/ai/backends/__init__.py,sha256=EAJRZS8G0DK7fffw_LHio9DkyYHwtzvz2Jo7AXk7pk4,303
 iints/ai/backends/base.py,sha256=BLgP03X-jebYkF9D5n5crawoPBmy3RSh4q3jaT8a9XM,274
 iints/ai/backends/mistral_api.py,sha256=dousHnzgzuik49822H8nCclYv5NoxHpTMLwtZPVj_TM,507
-iints/ai/backends/ollama.py,sha256=pH_UkrCmm6jI4_Hl-NM_7W4tXOwHZdZ0heeKph72NUk,8739
+iints/ai/backends/ollama.py,sha256=3VWi37ueh7oeNK6tPrq3Ks8gbfQhzgNlMHHnV_6BxAY,12189
 iints/analysis/__init__.py,sha256=Qx49KDy0deoxSnVORJB10_BsdezZXKsuoXTR0KZRcqg,411
 iints/analysis/algorithm_xray.py,sha256=-AtXkZsgnsiFQ_K-IozjIDWkq-dDn0i0zmqWVMhINP4,15952
 iints/analysis/baseline.py,sha256=PCFVb5vX0lYKChZvVk-8I_B5NLQQwGyx7Y6M3XjpIEY,3458
@@ -142,9 +142,9 @@ iints/validation/schemas.py,sha256=uXhiPxyfyvOgCA83ZPBIzlITOu663fWctYxOMXUyf1I,4
 iints/visualization/__init__.py,sha256=OdxVHDpY-9bDt8DTWWd-dspn1p0O9T908Cck-IGFaiM,640
 iints/visualization/cockpit.py,sha256=Y7hoJXcTEWQ8yLiU5X5abT58uqGGsQllftXJwqerG1E,25057
 iints/visualization/uncertainty_cloud.py,sha256=I5nNzSitgai21rkul31YNtJriSEmCeTsW0GWW2HUskY,19848
-iints_sdk_python35-1.1.2.dist-info/licenses/LICENSE,sha256=b1luljj2mWWDW10t_qFIqd9Z6euXAcDBmIXowWuUlm4,1417
-iints_sdk_python35-1.1.2.dist-info/METADATA,sha256=F-4k2b--KpKAGQpS6Kw8LP_AW3A1zzqATS7lgs_Mx8I,8887
-iints_sdk_python35-1.1.2.dist-info/WHEEL,sha256=aeYiig01lYGDzBgS8HxWXOg3uV61G9ijOsup-k9o1sk,91
-iints_sdk_python35-1.1.2.dist-info/entry_points.txt,sha256=aVioeLytTHG7WM7L3LIZ6XDJCKiSfqG-nVUQDVHPpQk,578
-iints_sdk_python35-1.1.2.dist-info/top_level.txt,sha256=7Usr6NQKiC9SpNFyCis81MmgXy71lDCr5unR8BNXZ0E,6
-iints_sdk_python35-1.1.2.dist-info/RECORD,,
+iints_sdk_python35-1.1.3.dist-info/licenses/LICENSE,sha256=b1luljj2mWWDW10t_qFIqd9Z6euXAcDBmIXowWuUlm4,1417
+iints_sdk_python35-1.1.3.dist-info/METADATA,sha256=5gDxBH6JpAOqvlisyv6aOye2tS4YRLBuybhz2-Oq7H0,9188
+iints_sdk_python35-1.1.3.dist-info/WHEEL,sha256=aeYiig01lYGDzBgS8HxWXOg3uV61G9ijOsup-k9o1sk,91
+iints_sdk_python35-1.1.3.dist-info/entry_points.txt,sha256=aVioeLytTHG7WM7L3LIZ6XDJCKiSfqG-nVUQDVHPpQk,578
+iints_sdk_python35-1.1.3.dist-info/top_level.txt,sha256=7Usr6NQKiC9SpNFyCis81MmgXy71lDCr5unR8BNXZ0E,6
+iints_sdk_python35-1.1.3.dist-info/RECORD,,