File size: 17,092 Bytes
4f7d15b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
# -----------------------------
# Single-file Chainlit app with inline "agents" shim
# Project: Multimodal Biomedical Imaging Tutor (education only)
# -----------------------------
import os, json
from dataclasses import dataclass, field
from typing import Any, Callable, Dict, List, Optional
from dotenv import load_dotenv
from pydantic import BaseModel, Field
import chainlit as cl
from openai import AsyncOpenAI as _SDKAsyncOpenAI

# =============================
# Inline "agents" shim
# =============================
def set_tracing_disabled(disabled: bool = True):
    return disabled

def function_tool(func: Callable):
    func._is_tool = True
    return func

def handoff(*args, **kwargs):
    return None

class InputGuardrail:
    def __init__(self, guardrail_function: Callable):
        self.guardrail_function = guardrail_function

@dataclass
class GuardrailFunctionOutput:
    output_info: Any
    tripwire_triggered: bool = False
    tripwire_message: str = ""

class InputGuardrailTripwireTriggered(Exception):
    pass

class AsyncOpenAI:
    def __init__(self, api_key: str, base_url: Optional[str] = None):
        kwargs = {"api_key": api_key}
        if base_url:
            kwargs["base_url"] = base_url
        self._client = _SDKAsyncOpenAI(**kwargs)

    @property
    def client(self):
        return self._client

class OpenAIChatCompletionsModel:
    def __init__(self, model: str, openai_client: AsyncOpenAI):
        self.model = model
        self.client = openai_client.client

@dataclass
class Agent:
    name: str
    instructions: str
    model: OpenAIChatCompletionsModel
    tools: Optional[List[Callable]] = field(default_factory=list)
    handoff_description: Optional[str] = None
    output_type: Optional[type] = None  # optional Pydantic model class
    input_guardrails: Optional[List[InputGuardrail]] = field(default_factory=list)

    def tool_specs(self) -> List[Dict[str, Any]]:
        specs = []
        for t in (self.tools or []):
            if getattr(t, "_is_tool", False):
                specs.append({
                    "type": "function",
                    "function": {
                        "name": t.__name__,
                        "description": (t.__doc__ or "")[:512],
                        "parameters": {
                            "type": "object",
                            "properties": {
                                p: {"type": "string"}
                                for p in t.__code__.co_varnames[:t.__code__.co_argcount]
                            },
                            "required": list(t.__code__.co_varnames[:t.__code__.co_argcount]),
                        },
                    },
                })
        return specs

class Runner:
    @staticmethod
    async def run(agent: Agent, user_input: str, context: Optional[Dict[str, Any]] = None):
        msgs = [
            {"role": "system", "content": agent.instructions},
            {"role": "user", "content": user_input},
        ]
        tools = agent.tool_specs()
        tool_map = {t.__name__: t for t in (agent.tools or []) if getattr(t, "_is_tool", False)}

        # simple tool loop
        for _ in range(4):
            resp = await agent.model.client.chat.completions.create(
                model=agent.model.model,
                messages=msgs,
                tools=tools if tools else None,
                tool_choice="auto" if tools else None,
            )

            choice = resp.choices[0]
            msg = choice.message
            msgs.append({"role": "assistant", "content": msg.content or "", "tool_calls": msg.tool_calls})

            if msg.tool_calls:
                for call in msg.tool_calls:
                    fn_name = call.function.name
                    args = json.loads(call.function.arguments or "{}")
                    if fn_name in tool_map:
                        try:
                            result = tool_map[fn_name](**args)
                        except Exception as e:
                            result = {"error": str(e)}
                    else:
                        result = {"error": f"Unknown tool: {fn_name}"}
                    msgs.append({
                        "role": "tool",
                        "tool_call_id": call.id,
                        "name": fn_name,
                        "content": json.dumps(result),
                    })
                continue  # let the model use tool outputs

            # finalize
            final_text = msg.content or ""
            final_obj = type("Result", (), {})()
            final_obj.final_output = final_text
            final_obj.context = context or {}
            if agent.output_type and issubclass(agent.output_type, BaseModel):
                try:
                    data = agent.output_type.model_validate_json(final_text)
                    final_obj.final_output = data.model_dump_json()
                    final_obj.final_output_as = lambda t: data
                except Exception:
                    final_obj.final_output_as = lambda t: final_text
            else:
                final_obj.final_output_as = lambda t: final_text
            return final_obj

        final_obj = type("Result", (), {})()
        final_obj.final_output = "Sorry, I couldn't complete the request."
        final_obj.context = context or {}
        final_obj.final_output_as = lambda t: final_obj.final_output
        return final_obj

# =============================
# App configuration
# =============================
load_dotenv()
API_KEY = os.environ.get("GEMINI_API_KEY") or os.environ.get("OPENAI_API_KEY")
if not API_KEY:
    raise RuntimeError(
        "Missing GEMINI_API_KEY (or OPENAI_API_KEY). "
        "Add it in the Space secrets or a .env file."
    )

set_tracing_disabled(True)

external_client: AsyncOpenAI = AsyncOpenAI(
    api_key=API_KEY,
    base_url="https://generativelanguage.googleapis.com/v1beta/openai/",
)
llm_model: OpenAIChatCompletionsModel = OpenAIChatCompletionsModel(
    model="gemini-2.5-flash",
    openai_client=external_client,
)

# =============================
# Domain models for tutor
# =============================
class Section(BaseModel):
    title: str
    bullets: List[str]

class TutorResponse(BaseModel):
    modality: str
    acquisition_overview: Section
    common_artifacts: Section
    preprocessing_methods: Section
    study_tips: Section
    caution: str

# =============================
# Tools
# =============================
@function_tool
def infer_modality_from_filename(filename: str) -> dict:
    """
    Guess modality (MRI/X-ray/CT/Ultrasound) from filename keywords.
    Returns: {"modality": "<guess or unknown>"}
    """
    f = (filename or "").lower()
    guess = "unknown"
    mapping = {
        "xray": "X-ray", "x_ray": "X-ray", "xr": "X-ray", "cxr": "X-ray",
        "mri": "MRI", "t1": "MRI", "t2": "MRI", "flair": "MRI", "dwi": "MRI", "adc": "MRI",
        "ct": "CT", "cta": "CT",
        "ultrasound": "Ultrasound", "usg": "Ultrasound", "echo": "Ultrasound",
    }
    for key, mod in mapping.items():
        if key in f:
            guess = mod
            break
    return {"modality": guess}

@function_tool
def imaging_reference_guide(modality: str) -> dict:
    """
    Educational points for acquisition, artifacts, preprocessing, and study tips by modality.
    Education only (no diagnosis).
    """
    mod = (modality or "").strip().lower()
    if mod in ["xray", "x-ray", "x_ray"]:
        return {
            "acquisition": [
                "Projection radiography using ionizing radiation.",
                "Common views: AP, PA, lateral; exposure (kVp/mAs) and positioning matter.",
                "Grids/collimation reduce scatter and improve contrast."
            ],
            "artifacts": [
                "Motion blur; under/overexposure affecting contrast.",
                "Grid cut-off; foreign objects (buttons, jewelry).",
                "Magnification/distortion from object–detector distance."
            ],
            "preprocessing": [
                "Denoising (median/NLM), histogram equalization.",
                "Window/level selection (bone vs soft tissue) for teaching.",
                "Edge enhancement (unsharp mask) with caution (halo artifacts)."
            ],
            "study_tips": [
                "Use a systematic approach (e.g., ABCDE for chest X-ray).",
                "Compare sides; verify devices, labels, positioning.",
                "Correlate with clinical scenario; keep a checklist."
            ],
        }
    if mod in ["mri", "mr"]:
        return {
            "acquisition": [
                "MR uses RF pulses in a strong magnetic field; sequences set contrast.",
                "Key sequences: T1, T2, FLAIR, DWI/ADC, GRE/SWI.",
                "TR/TE/flip angle shape SNR, contrast, time."
            ],
            "artifacts": [
                "Motion/ghosting (movement, pulsation).",
                "Susceptibility (metal, air-bone interfaces).",
                "Chemical shift, Gibbs ringing.",
                "B0/B1 inhomogeneity causing intensity bias."
            ],
            "preprocessing": [
                "Bias-field correction (N4).",
                "Denoising (non-local means), registration/normalization.",
                "Skull stripping (brain), intensity standardization."
            ],
            "study_tips": [
                "Know sequence intent (T1 anatomy, T2 fluid, FLAIR edema).",
                "Check diffusion for acute ischemia (with ADC).",
                "Use consistent windowing for longitudinal comparison."
            ],
        }
    if mod == "ct":
        return {
            "acquisition": [
                "Helical CT reconstructs attenuation in Hounsfield Units.",
                "Kernels (bone vs soft) change sharpness/noise.",
                "Contrast phases (arterial/venous) match the task."
            ],
            "artifacts": [
                "Beam hardening (streaks), partial volume.",
                "Motion (breathing/cardiac).",
                "Metal artifacts; consider MAR algorithms."
            ],
            "preprocessing": [
                "Denoising (bilateral/NLM) while preserving edges.",
                "Appropriate window/level (lung, mediastinum, bone).",
                "Iterative reconstruction / metal artifact reduction."
            ],
            "study_tips": [
                "Use standard planes; scroll systematically.",
                "Compare windows; document sizes/HU as needed.",
                "Correlate phase with the clinical question."
            ],
        }
    return {
        "acquisition": [
            "Acquisition parameters define contrast, resolution, and noise.",
            "Positioning and motion control are crucial for quality."
        ],
        "artifacts": [
            "Motion blur/ghosting; foreign objects and hardware.",
            "Parameter misconfiguration harms interpretability."
        ],
        "preprocessing": [
            "Denoising and contrast normalization for clarity.",
            "Registration to standard planes for comparison."
        ],
        "study_tips": [
            "Adopt a checklist; compare across time or sides.",
            "Learn modality-specific knobs (window/level, sequences)."
        ],
    }

@function_tool
def file_facts(filename: str, size_bytes: str) -> dict:
    """
    Returns lightweight file facts: filename and byte size (as string).
    """
    try:
        size = int(size_bytes)
    except Exception:
        size = -1
    return {"filename": filename, "size_bytes": size}

# =============================
# Agents
# =============================
tutor_instructions = (
    "You are a Biomedical Imaging Education Tutor. TEACH, do not diagnose.\n"
    "Given an uploaded MRI or X-ray, provide:\n"
    "1) Acquisition overview\n"
    "2) Common artifacts\n"
    "3) Preprocessing methods\n"
    "4) Study tips\n"
    "5) A caution line: education only, no diagnosis\n"
    "Use tools to infer modality from filename and to fetch a modality reference guide.\n"
    "If unclear, provide a generic overview and ask for clarification.\n"
    "Always respond as concise, well-structured bullet points.\n"
    "Absolutely avoid clinical diagnosis, disease identification, or treatment advice."
)

tutor_agent = Agent(
    name="Biomedical Imaging Tutor",
    instructions=tutor_instructions,
    model=llm_model,
    tools=[infer_modality_from_filename, imaging_reference_guide, file_facts],
)

class SafetyCheck(BaseModel):
    unsafe_medical_advice: bool
    requests_diagnosis: bool
    pii_included: bool
    reasoning: str

guardrail_agent = Agent(
    name="Safety Classifier",
    instructions=(
        "Classify if the user's message requests medical diagnosis or unsafe medical advice, "
        "and if it includes personal identifiers. Respond as JSON with fields: "
        "{unsafe_medical_advice: bool, requests_diagnosis: bool, pii_included: bool, reasoning: string}."
    ),
    model=llm_model,
)

# =============================
# Chainlit flows
# =============================
WELCOME = (
    "🎓 **Multimodal Biomedical Imaging Tutor**\n\n"
    "Upload an **MRI** or **X-ray** image (PNG/JPG). I’ll explain:\n"
    "• Acquisition (how it’s made)\n"
    "• Common artifacts (what to watch for)\n"
    "• Preprocessing for study/teaching\n\n"
    "⚠️ *Education only — I do not provide diagnosis. For clinical concerns, consult a professional.*"
)

@cl.on_chat_start
async def on_chat_start():
    await cl.Message(content=WELCOME).send()
    files = await cl.AskFileMessage(
        content="Please upload an **MRI or X-ray** image (PNG/JPG).",
        accept=["image/png", "image/jpeg"],
        max_size_mb=15,
        max_files=1,
        timeout=180,
    ).send()

    if not files:
        await cl.Message(content="No file uploaded. You can still ask general imaging questions.").send()
        return

    f = files[0]
    cl.user_session.set("last_file_path", f.path)
    cl.user_session.set("last_file_name", f.name)
    cl.user_session.set("last_file_size", f.size)

    await cl.Message(
        content=f"Received **{f.name}** ({f.size} bytes). "
                "Ask: *“Explain acquisition & artifacts for this image.”*"
    ).send()

@cl.on_message
async def on_message(message: cl.Message):
    # Safety check
    try:
        safety = await Runner.run(guardrail_agent, message.content)
        # parse best-effort
        parsed = safety.final_output
        try:
            data = json.loads(parsed) if isinstance(parsed, str) else parsed
        except Exception:
            data = {}
        if isinstance(data, dict):
            if data.get("unsafe_medical_advice") or data.get("requests_diagnosis"):
                await cl.Message(
                    content=(
                        "🚫 I can’t provide medical diagnoses or treatment advice.\n"
                        "I’m happy to explain **imaging concepts**, **artifacts**, and **preprocessing** for learning."
                    )
                ).send()
                return
    except Exception:
        pass  # continue gracefully

    # Context from last upload
    file_name = cl.user_session.get("last_file_name")
    file_size = cl.user_session.get("last_file_size")

    context_note = ""
    if file_name:
        context_note += f"The user uploaded a file named '{file_name}'.\n"
    if file_size is not None:
        context_note += f"File size: {file_size} bytes.\n"

    user_query = message.content
    if context_note:
        user_query = f"{user_query}\n\n[Context]\n{context_note}"

    # Run tutor
    result = await Runner.run(tutor_agent, user_query)

    # Quick reference facts
    facts_md = ""
    try:
        modality = infer_modality_from_filename(file_name or "").get("modality", "unknown")
        guide = imaging_reference_guide(modality)
        acq = "\n".join([f"- {b}" for b in guide.get("acquisition", [])])
        art = "\n".join([f"- {b}" for b in guide.get("artifacts", [])])
        prep = "\n".join([f"- {b}" for b in guide.get("preprocessing", [])])
        tips = "\n".join([f"- {b}" for b in guide.get("study_tips", [])])

        facts_md = (
            f"### 📁 File\n"
            f"- Name: `{file_name or 'unknown'}`\n"
            f"- Size: `{file_size if file_size is not None else 'unknown'} bytes`\n\n"
            f"### 🔎 Modality (guess)\n- {modality}\n\n"
            f"### 📚 Reference Guide (study)\n"
            f"**Acquisition**\n{acq or '- (general)'}\n\n"
            f"**Common Artifacts**\n{art or '- (general)'}\n\n"
            f"**Preprocessing Ideas**\n{prep or '- (general)'}\n\n"
            f"**Study Tips**\n{tips or '- (general)'}\n\n"
            f"> ⚠️ Education only — no diagnosis.\n"
        )
    except Exception:
        pass

    text = result.final_output or "I couldn’t generate an explanation."
    await cl.Message(content=f"{facts_md}\n---\n{text}").send()