iints-sdk-python35 1.1.2__tar.gz → 1.2.0__tar.gz

{iints_sdk_python35-1.1.2/src/iints_sdk_python35.egg-info → iints_sdk_python35-1.2.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: iints-sdk-python35
-Version: 1.1.2
+Version: 1.2.0
 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
@@ -81,6 +81,47 @@ cd iints_quickstart
 iints presets run --name baseline_t1d --algo algorithms/example_algorithm.py
 ```
 
+## CareLink Import
+
+The SDK can now ingest Medtronic CareLink / MiniMed CSV exports and convert them into the standard IINTS schema:
+
+```bash
+iints import-carelink \
+  --input-csv "/path/to/CareLink export.csv" \
+  --output-dir results/imported_carelink
+```
+
+This writes:
+- `cgm_standard.csv`
+- `scenario.json`
+- `carelink_summary.json`
+
+It extracts glucose, carb, and insulin events from the CareLink event log and aligns them onto an IINTS-ready timeline.
+
+If you want a reusable personal-data workspace in one command, build a CareLink workbench:
+
+```bash
+iints carelink-workbench \
+  --input-csv "/path/to/CareLink export.csv" \
+  --output-dir results/personal_carelink
+```
+
+This adds:
+- `carelink_timeline.csv`
+- `carelink_metrics.json`
+- `carelink_dashboard.png`
+- `carelink_poster.png`
+- `carelink_dashboard.html`
+- `ai/report_payload.json`
+- `ai/trends_payload.json`
+- `ai/anomalies_payload.json`
+- `ai/step_riskiest.json`
+
+That workbench is designed for three things:
+- inspect your own data visually
+- reuse the generated `scenario.json` inside IINTS experiments
+- let the local AI assistant explain what the imported patterns mean
+
 ## AI Assistant (Ministral 3 Open-Weight via Ollama)
 
 The SDK now includes a research-only AI assistant layer for explanations and run summaries.
@@ -110,6 +151,8 @@ 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
@@ -120,6 +163,18 @@ iints ai prepare results/<run_id>
 iints ai report results/<run_id>
 ```
 
+For imported CareLink data, the matching flow is:
+
+```bash
+iints carelink-workbench \
+  --input-csv "/path/to/CareLink export.csv" \
+  --output-dir results/personal_carelink
+
+iints ai report results/personal_carelink --model ministral-3:3b
+iints ai trends results/personal_carelink --model ministral-3:3b
+iints ai explain results/personal_carelink --model ministral-3:3b
+```
+
 Direct JSON mode still works if you already have your own payloads and signed MDMP artifact:
 
 ```bash
@@ -134,9 +189,35 @@ 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/`.
+- `iints carelink-workbench` now does the same kind of AI preparation for imported personal CareLink data and also generates a dashboard PNG/HTML pair.
+- 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.
+- After `iints carelink-workbench`, you can point those same AI commands directly at the generated CareLink workspace directory.
 - Output is research-only and not medical advice.
 
+## Jury Poster / Demo Graphic
+
+You can now generate a poster-style PNG directly from one to three real run bundles:
+
+```bash
+iints poster \
+  --run-dir results/normal_run \
+  --run-dir results/meal_stress \
+  --run-dir results/supervisor_override \
+  --label "Normal Run" \
+  --label "Meal Stress Test" \
+  --label "Supervisor Override" \
+  --output-path results/posters/iints_results_poster.png
+```
+
+The poster shows:
+- glucose curves with the target range highlighted
+- meal events
+- supervisor interventions
+- panel summaries with TIR, hypo time, meal count, and intervention count
+
+If you omit `--run-dir`, the CLI auto-discovers the latest run bundles under `./results`.
+
 Troubleshooting:
 - If `iints ai ...` says `No such command 'ai'`, your environment usually still has a legacy `iints` package installed alongside `iints-sdk-python35`.
 - Run `iints-sdk-doctor` first.
@@ -144,7 +225,7 @@ 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]==1.2.0"
 hash -r
 ```
 

{iints_sdk_python35-1.1.2 → iints_sdk_python35-1.2.0}/README.md

@@ -30,6 +30,47 @@ cd iints_quickstart
 iints presets run --name baseline_t1d --algo algorithms/example_algorithm.py
 ```
 
+## CareLink Import
+
+The SDK can now ingest Medtronic CareLink / MiniMed CSV exports and convert them into the standard IINTS schema:
+
+```bash
+iints import-carelink \
+  --input-csv "/path/to/CareLink export.csv" \
+  --output-dir results/imported_carelink
+```
+
+This writes:
+- `cgm_standard.csv`
+- `scenario.json`
+- `carelink_summary.json`
+
+It extracts glucose, carb, and insulin events from the CareLink event log and aligns them onto an IINTS-ready timeline.
+
+If you want a reusable personal-data workspace in one command, build a CareLink workbench:
+
+```bash
+iints carelink-workbench \
+  --input-csv "/path/to/CareLink export.csv" \
+  --output-dir results/personal_carelink
+```
+
+This adds:
+- `carelink_timeline.csv`
+- `carelink_metrics.json`
+- `carelink_dashboard.png`
+- `carelink_poster.png`
+- `carelink_dashboard.html`
+- `ai/report_payload.json`
+- `ai/trends_payload.json`
+- `ai/anomalies_payload.json`
+- `ai/step_riskiest.json`
+
+That workbench is designed for three things:
+- inspect your own data visually
+- reuse the generated `scenario.json` inside IINTS experiments
+- let the local AI assistant explain what the imported patterns mean
+
 ## AI Assistant (Ministral 3 Open-Weight via Ollama)
 
 The SDK now includes a research-only AI assistant layer for explanations and run summaries.
@@ -59,6 +100,8 @@ 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
@@ -69,6 +112,18 @@ iints ai prepare results/<run_id>
 iints ai report results/<run_id>
 ```
 
+For imported CareLink data, the matching flow is:
+
+```bash
+iints carelink-workbench \
+  --input-csv "/path/to/CareLink export.csv" \
+  --output-dir results/personal_carelink
+
+iints ai report results/personal_carelink --model ministral-3:3b
+iints ai trends results/personal_carelink --model ministral-3:3b
+iints ai explain results/personal_carelink --model ministral-3:3b
+```
+
 Direct JSON mode still works if you already have your own payloads and signed MDMP artifact:
 
 ```bash
@@ -83,9 +138,35 @@ 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/`.
+- `iints carelink-workbench` now does the same kind of AI preparation for imported personal CareLink data and also generates a dashboard PNG/HTML pair.
+- 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.
+- After `iints carelink-workbench`, you can point those same AI commands directly at the generated CareLink workspace directory.
 - Output is research-only and not medical advice.
 
+## Jury Poster / Demo Graphic
+
+You can now generate a poster-style PNG directly from one to three real run bundles:
+
+```bash
+iints poster \
+  --run-dir results/normal_run \
+  --run-dir results/meal_stress \
+  --run-dir results/supervisor_override \
+  --label "Normal Run" \
+  --label "Meal Stress Test" \
+  --label "Supervisor Override" \
+  --output-path results/posters/iints_results_poster.png
+```
+
+The poster shows:
+- glucose curves with the target range highlighted
+- meal events
+- supervisor interventions
+- panel summaries with TIR, hypo time, meal count, and intervention count
+
+If you omit `--run-dir`, the CLI auto-discovers the latest run bundles under `./results`.
+
 Troubleshooting:
 - If `iints ai ...` says `No such command 'ai'`, your environment usually still has a legacy `iints` package installed alongside `iints-sdk-python35`.
 - Run `iints-sdk-doctor` first.
@@ -93,7 +174,7 @@ 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]==1.2.0"
 hash -r
 ```
 

{iints_sdk_python35-1.1.2 → iints_sdk_python35-1.2.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "iints-sdk-python35" 
-version = "1.1.2"
+version = "1.2.0"
 authors = [
   { name="Rune Bobbaers", email="rune.bobbaers@gmail.com" },
 ]

{iints_sdk_python35-1.1.2 → iints_sdk_python35-1.2.0}/src/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__ = "1.2.0"
 
 # 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.
@@ -56,17 +56,23 @@ from .data.importer import (
     export_demo_csv,
     export_standard_csv,
     guess_column_mapping,
+    import_carelink_csv,
+    import_carelink_timeline,
     import_cgm_csv,
     import_cgm_dataframe,
+    load_carelink_event_log,
     load_demo_dataframe,
     scenario_from_csv,
     scenario_from_dataframe,
+    summarize_carelink_csv,
 )
 from .data.nightscout import NightscoutConfig, import_nightscout
 from .data.tidepool import TidepoolClient, load_openapi_spec
 from .data.guardians import mdmp_gate, MDMPGateError
 from .data.synthetic_mirror import generate_synthetic_mirror, SyntheticMirrorArtifact
 from .analysis.metrics import generate_benchmark_metrics # Added for benchmark
+from .analysis.carelink_workbench import build_carelink_workbench
+from .analysis.poster import generate_results_poster
 from .analysis.reporting import ClinicalReportGenerator
 from .analysis.edge_efficiency import EnergyEstimate, estimate_energy_per_decision
 from .ai import AIResponse, IINTSAssistant, MDMPGuard
@@ -160,11 +166,15 @@ __all__ = [
     "export_demo_csv",
     "export_standard_csv",
     "guess_column_mapping",
+    "import_carelink_csv",
+    "import_carelink_timeline",
     "import_cgm_csv",
     "import_cgm_dataframe",
+    "load_carelink_event_log",
     "load_demo_dataframe",
     "scenario_from_csv",
     "scenario_from_dataframe",
+    "summarize_carelink_csv",
     "NightscoutConfig",
     "import_nightscout",
     "TidepoolClient",
@@ -175,6 +185,7 @@ __all__ = [
     "SyntheticMirrorArtifact",
     # Analysis Metrics
     "generate_benchmark_metrics",
+    "build_carelink_workbench",
     "ClinicalReportGenerator",
     "EnergyEstimate",
     "estimate_energy_per_decision",
@@ -185,6 +196,7 @@ __all__ = [
     "generate_report",
     "generate_quickstart_report",
     "generate_demo_report",
+    "generate_results_poster",
     # High-level API
     "run_simulation",
     "run_full",

{iints_sdk_python35-1.1.2 → iints_sdk_python35-1.2.0}/src/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_sdk_python35-1.1.2 → iints_sdk_python35-1.2.0}/src/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 → iints_sdk_python35-1.2.0}/src/iints/ai/prompts.py

@@ -8,8 +8,9 @@ TaskName = Literal["explain_decision", "analyze_trends", "detect_anomalies", "ge
 MAX_PROMPT_PAYLOAD_CHARS = 12000
 
 SYSTEM_PROMPT = (
-    "You are the IINTS-AF research assistant for closed-loop insulin delivery simulations. "
-    "Explain simulation behavior clearly, conservatively, and in plain language. "
+    "You are the IINTS-AF research assistant for closed-loop insulin delivery simulations "
+    "and imported glucose datasets. "
+    "Explain glycemic behavior clearly, conservatively, and in plain language. "
     "Do not give medical advice, treatment instructions, or patient-specific recommendations. "
     "State uncertainty when the input is incomplete. "
     "For research use only."
@@ -17,39 +18,39 @@ SYSTEM_PROMPT = (
 
 TASK_TEMPLATES: dict[TaskName, str] = {
     "explain_decision": (
-        "Given this single simulation step, explain:\n"
-        "1. What the algorithm decided and why\n"
-        "2. Whether the independent safety supervisor likely intervened\n"
-        "3. Whether the decision appears safe in context\n\n"
+        "Given this single decision step or noteworthy glucose snapshot, explain:\n"
+        "1. What is happening in the data and why it stands out\n"
+        "2. Whether there are safety signals, supervision, or notable context clues\n"
+        "3. What a research user should pay attention to next\n\n"
         "Respond in 3 short paragraphs.\n\n"
-        "Simulation step JSON:\n{data}"
+        "Input JSON:\n{data}"
     ),
     "analyze_trends": (
-        "Review this glucose-oriented simulation payload and summarize the main glycemic trends.\n"
+        "Review this glucose-oriented payload and summarize the main glycemic trends.\n"
         "Focus on direction, stability, excursions, and likely triggers.\n"
         "Respond with:\n"
         "- Trend summary\n"
         "- Main risk signals\n"
         "- Short operational takeaway\n\n"
-        "Simulation payload JSON:\n{data}"
+        "Payload JSON:\n{data}"
     ),
     "detect_anomalies": (
-        "Inspect this run summary and identify unusual patterns, inconsistent values, or clinically relevant anomalies.\n"
+        "Inspect this run or imported-data summary and identify unusual patterns, inconsistent values, or clinically relevant anomalies.\n"
         "Respond with:\n"
         "- Detected anomalies\n"
         "- Why each anomaly matters\n"
         "- Whether follow-up validation is recommended\n\n"
-        "Run summary JSON:\n{data}"
+        "Summary JSON:\n{data}"
     ),
     "generate_report": (
-        "Write a concise markdown report for this IINTS-AF simulation run.\n"
+        "Write a concise markdown report for this IINTS-AF simulation run or imported personal glucose dataset.\n"
         "Include sections:\n"
         "1. Executive summary\n"
         "2. Glycemic behavior\n"
-        "3. Safety and supervisor behavior\n"
-        "4. Notable events or anomalies\n"
+        "3. Safety, supervision, or device behavior\n"
+        "4. Notable events, patterns, or anomalies\n"
         "5. Research-only conclusion\n\n"
-        "Simulation run JSON:\n{data}"
+        "Input JSON:\n{data}"
     ),
 }
 

{iints_sdk_python35-1.1.2 → iints_sdk_python35-1.2.0}/src/iints/analysis/__init__.py

@@ -1,12 +1,16 @@
 from .clinical_metrics import ClinicalMetricsCalculator, ClinicalMetricsResult
 from .baseline import compute_metrics, run_baseline_comparison, write_baseline_comparison
+from .carelink_workbench import build_carelink_workbench
+from .poster import generate_results_poster
 from .reporting import ClinicalReportGenerator
 
 __all__ = [
+    "build_carelink_workbench",
     "ClinicalMetricsCalculator",
     "ClinicalMetricsResult",
     "ClinicalReportGenerator",
     "compute_metrics",
+    "generate_results_poster",
     "run_baseline_comparison",
     "write_baseline_comparison",
 ]