update: two stages processing

app.py

@@ -1,8 +1,11 @@
 import logging
 import os
 import tempfile
+import time
+from dataclasses import dataclass
 from pathlib import Path
 from typing import Optional
+from uuid import uuid4
 
 from fastapi import BackgroundTasks, FastAPI, File, Form, HTTPException, UploadFile
 from fastapi.middleware.cors import CORSMiddleware
@@ -10,6 +13,7 @@ from fastapi.responses import FileResponse, HTMLResponse, JSONResponse
 import uvicorn
 
 from inference import run_inference
+from mission_planner import MissionPlan, get_mission_plan
 
 logging.basicConfig(level=logging.INFO)
 
@@ -50,21 +54,94 @@ def _schedule_cleanup(background_tasks: BackgroundTasks, path: str) -> None:
     background_tasks.add_task(_cleanup)
 
 
-def _validate_inputs(video: UploadFile | None, prompt: str | None) -> None:
+@dataclass
+class CachedMission:
+    prompt: str
+    detector: Optional[str]
+    plan: MissionPlan
+    created_at: float
+
+
+MISSION_CACHE: dict[str, CachedMission] = {}
+MISSION_CACHE_TTL_SECONDS = 3600.0
+
+
+def _prune_mission_cache() -> None:
+    now = time.time()
+    expired = [
+        key
+        for key, entry in MISSION_CACHE.items()
+        if now - entry.created_at > MISSION_CACHE_TTL_SECONDS
+    ]
+    for key in expired:
+        MISSION_CACHE.pop(key, None)
+
+
+def _store_mission_plan(prompt: str, detector: Optional[str], plan: MissionPlan) -> str:
+    _prune_mission_cache()
+    mission_id = uuid4().hex
+    MISSION_CACHE[mission_id] = CachedMission(
+        prompt=prompt,
+        detector=detector,
+        plan=plan,
+        created_at=time.time(),
+    )
+    return mission_id
+
+
+def _get_cached_mission(mission_id: str) -> CachedMission:
+    _prune_mission_cache()
+    entry = MISSION_CACHE.get(mission_id)
+    if entry is None:
+        raise HTTPException(status_code=404, detail="Mission prompt not found. Please set it again.")
+    return entry
+
+
+def _resolve_mission_plan(prompt: Optional[str], mission_id: Optional[str]) -> tuple[MissionPlan, str]:
+    if mission_id:
+        cached = _get_cached_mission(mission_id)
+        return cached.plan, cached.prompt
+    normalized_prompt = (prompt or "").strip()
+    if not normalized_prompt:
+        raise HTTPException(status_code=400, detail="Mission prompt is required.")
+    plan = get_mission_plan(normalized_prompt)
+    return plan, normalized_prompt
+
+
+def _validate_inputs(video: UploadFile | None, prompt: str | None, mission_id: str | None) -> None:
     if video is None:
         raise HTTPException(status_code=400, detail="Video file is required.")
-    if not prompt:
+    if not prompt and not mission_id:
+        raise HTTPException(status_code=400, detail="Mission prompt is required.")
+
+
+@app.post("/mission_plan")
+async def mission_plan_endpoint(
+    prompt: str = Form(...),
+    detector: Optional[str] = Form(None),
+):
+    normalized_prompt = (prompt or "").strip()
+    if not normalized_prompt:
         raise HTTPException(status_code=400, detail="Prompt is required.")
+    try:
+        plan = get_mission_plan(normalized_prompt)
+    except Exception as exc:
+        logging.exception("Mission planning failed.")
+        raise HTTPException(status_code=500, detail=str(exc))
+
+    mission_id = _store_mission_plan(normalized_prompt, detector, plan)
+    return {"mission_id": mission_id, "mission_plan": plan.to_dict()}
 
 
 @app.post("/process_video")
 async def process_video(
     background_tasks: BackgroundTasks,
     video: UploadFile = File(...),
-    prompt: str = Form(...),
+    prompt: Optional[str] = Form(None),
+    mission_id: Optional[str] = Form(None),
     detector: Optional[str] = Form(None),
 ):
-    _validate_inputs(video, prompt)
+    _validate_inputs(video, prompt, mission_id)
 
     try:
         input_path = _save_upload_to_tmp(video)
@@ -77,13 +154,15 @@ async def process_video(
     fd, output_path = tempfile.mkstemp(prefix="output_", suffix=".mp4", dir="/tmp")
     os.close(fd)
 
+    mission_plan, mission_prompt = _resolve_mission_plan(prompt, mission_id)
     try:
         output_path, _, _ = run_inference(
             input_path,
             output_path,
-            prompt,
+            mission_prompt,
             detector_name=detector,
             generate_summary=False,
+            mission_plan=mission_plan,
         )
     except ValueError as exc:
         logging.exception("Video decoding failed.")
@@ -110,10 +189,12 @@ async def process_video(
 @app.post("/mission_summary")
 async def mission_summary(
     video: UploadFile = File(...),
-    prompt: str = Form(...),
+    prompt: Optional[str] = Form(None),
+    mission_id: Optional[str] = Form(None),
     detector: Optional[str] = Form(None),
 ):
-    _validate_inputs(video, prompt)
+    _validate_inputs(video, prompt, mission_id)
+    mission_plan, mission_prompt = _resolve_mission_plan(prompt, mission_id)
     try:
         input_path = _save_upload_to_tmp(video)
     except Exception:
@@ -123,13 +204,14 @@ async def mission_summary(
         await video.close()
 
     try:
-        _, mission_plan, mission_summary = run_inference(
+        _, _, mission_summary = run_inference(
             input_path,
             output_video_path=None,
-            mission_prompt=prompt,
+            mission_prompt=mission_prompt,
             detector_name=detector,
             write_output_video=False,
             generate_summary=True,
+            mission_plan=mission_plan,
         )
     except ValueError as exc:
         logging.exception("Video decoding failed.")

demo.html

@@ -22,8 +22,8 @@ h1 {
 }
 
 .container {
-    width: 80%;
-    max-width: 900px;
+    width: 90%;
+    max-width: 1100px;
     margin: auto;
     background: #1e1e1e;
     padding: 25px;
@@ -31,28 +31,39 @@ h1 {
     box-shadow: 0 15px 40px rgba(0, 0, 0, 0.25);
 }
 
+.form-grid {
+    display: flex;
+    flex-wrap: wrap;
+    gap: 20px;
+    margin-top: 10px;
+}
+
+.form-control {
+    flex: 1;
+    min-width: 260px;
+}
+
 .label {
     font-size: 15px;
     font-weight: 600;
-    margin-top: 18px;
     display: block;
     color: #bdbdbd;
 }
 
 input[type="text"] {
-    width: 100%;
     padding: 12px;
-    margin-top: 5px;
     border: 1px solid #2c2c2c;
     background-color: #161616;
     color: #fff;
     font-size: 15px;
     border-radius: 8px;
+    width: 100%;
+    margin-top: 5px;
 }
 
 input[type="file"],
 select {
-    margin-top: 10px;
+    margin-top: 5px;
     padding: 10px;
     background-color: #161616;
     border: 1px solid #2c2c2c;
@@ -79,6 +90,30 @@ button:hover {
     background-color: #4f4f4f;
 }
 
+button:disabled {
+    background-color: #1f1f1f;
+    color: #6e6e6e;
+    cursor: not-allowed;
+}
+
+.secondary-button {
+    width: auto;
+    padding: 10px 18px;
+    margin-top: 10px;
+    background-color: #262626;
+}
+
+.secondary-button:hover {
+    background-color: #3a3a3a;
+}
+
+.prompt-actions {
+    display: flex;
+    flex-wrap: wrap;
+    gap: 10px;
+    margin-top: 8px;
+}
+
 .output-section {
     margin-top: 35px;
     padding: 20px;
@@ -99,13 +134,39 @@ button:hover {
     border-radius: 8px;
 }
 
-#processedVideo {
+.video-section {
     margin-top: 20px;
+    display: flex;
+    flex-wrap: wrap;
+    gap: 25px;
+}
+
+.video-panel {
+    flex: 1;
+    min-width: 320px;
+}
+
+.video-panel h2 {
+    color: #00ffea;
+    margin-bottom: 10px;
+    font-size: 18px;
+}
+
+.mission-video {
     width: 100%;
+    max-height: 420px;
+    object-fit: contain;
     border: 1px solid #2a2a2a;
     border-radius: 8px;
     background-color: #000;
 }
+
+.status-message {
+    margin-top: 15px;
+    color: #ffa95c;
+    font-size: 13px;
+    min-height: 18px;
+}
 </style>
 
 </head>
@@ -115,11 +176,19 @@ button:hover {
 
 <div class="container">
 
-    <label class="label">MISSION PROMPT</label>
-    <input id="missionPrompt" type="text" placeholder="e.g., Track hostile drone movement...">
-
-    <label class="label">UPLOAD VIDEO (.mp4)</label>
-    <input id="videoInput" type="file" accept="video/mp4">
+    <div class="form-grid">
+        <div class="form-control">
+            <label class="label">MISSION PROMPT</label>
+            <input id="missionPrompt" type="text" placeholder="e.g., Track hostile drone movement...">
+            <div class="prompt-actions">
+                <button type="button" id="setMissionButton" class="secondary-button" onclick="stageMissionPrompt()">Set Mission Prompt</button>
+            </div>
+        </div>
+        <div class="form-control">
+            <label class="label">UPLOAD VIDEO (.mp4)</label>
+            <input id="videoInput" type="file" accept="video/mp4" disabled>
+        </div>
+    </div>
 
     <label class="label">OBJECT DETECTOR</label>
     <select id="detectorSelect">
@@ -128,14 +197,22 @@ button:hover {
         <option value="hf_yolov8_defence">YOLOv8m Defence</option>
     </select>
 
-    <button onclick="executeMission()">EXECUTE MISSION</button>
+    <button type="button" id="executeButton" onclick="executeMission()" disabled>EXECUTE MISSION</button>
 
     <div class="output-section">
-        <h2 style="color:#00ffea; margin-bottom:10px;">MISSION SUMMARY</h2>
+        <div class="video-section">
+            <div class="video-panel">
+                <h2>ORIGINAL VIDEO</h2>
+                <video id="originalVideo" class="mission-video" controls></video>
+            </div>
+            <div class="video-panel">
+                <h2>PROCESSED VIDEO FEED</h2>
+                <video id="processedVideo" class="mission-video" controls></video>
+            </div>
+        </div>
+        <h2 style="color:#00ffea; margin:25px 0 10px;">MISSION SUMMARY</h2>
         <div id="summary">(Awaiting mission results...)</div>
-        <h2 style="color:#00ffea; margin-top:25px;">PROCESSED VIDEO FEED</h2>
-        <video id="processedVideo" controls></video>
-        <div id="status" style="margin-top:15px;color:#ffa95c;font-size:13px;"></div>
+        <div id="status" class="status-message"></div>
     </div>
 
 </div>
@@ -144,27 +221,132 @@ button:hover {
 const API_BASE_URL = "https://biaslab2025-demo-2025.hf.space";
 const PROCESS_VIDEO_URL = `${API_BASE_URL}/process_video`;
 const SUMMARY_URL = `${API_BASE_URL}/mission_summary`;
+const MISSION_PLAN_URL = `${API_BASE_URL}/mission_plan`;
+
+const missionInputEl = document.getElementById("missionPrompt");
+const detectorSelectEl = document.getElementById("detectorSelect");
+const originalVideoEl = document.getElementById("originalVideo");
+const videoInputEl = document.getElementById("videoInput");
+const summaryEl = document.getElementById("summary");
+const statusEl = document.getElementById("status");
+const setMissionButton = document.getElementById("setMissionButton");
+const executeButton = document.getElementById("executeButton");
+
+let originalVideoUrl = null;
+let currentMissionId = null;
+let missionReady = false;
+let missionRequestPending = false;
+let videoProcessing = false;
+
+missionInputEl.addEventListener("input", handleMissionPromptEdit);
+
+videoInputEl.addEventListener("change", () => {
+    const uploaded = videoInputEl.files[0];
+    if (uploaded) {
+        setOriginalPreview(uploaded);
+    } else if (originalVideoUrl) {
+        URL.revokeObjectURL(originalVideoUrl);
+        originalVideoEl.removeAttribute("src");
+        originalVideoEl.load();
+        originalVideoUrl = null;
+    }
+});
+
+function handleMissionPromptEdit() {
+    if (!missionReady && !currentMissionId) {
+        return;
+    }
+    missionReady = false;
+    currentMissionId = null;
+    setStatus("Mission prompt changed. Set it again to process videos.");
+    updateControlState();
+}
+
+function updateControlState() {
+    const disableVideoActions = !missionReady || missionRequestPending || videoProcessing;
+    videoInputEl.disabled = disableVideoActions;
+    executeButton.disabled = disableVideoActions;
+    setMissionButton.disabled = missionRequestPending;
+}
+
+function setStatus(message, tone = "info") {
+    statusEl.textContent = message || "";
+    if (tone === "error") {
+        statusEl.style.color = "#ff7b7b";
+    } else if (tone === "success") {
+        statusEl.style.color = "#7bffb3";
+    } else {
+        statusEl.style.color = "#ffa95c";
+    }
+}
+
+async function stageMissionPrompt() {
+    const mission = missionInputEl.value.trim();
+    if (!mission) {
+        alert("Enter a mission prompt before setting it.");
+        return;
+    }
+    missionRequestPending = true;
+    missionReady = false;
+    currentMissionId = null;
+    updateControlState();
+    setStatus("Preparing mission prompt...");
+    try {
+        const form = new FormData();
+        form.append("prompt", mission);
+        form.append("detector", detectorSelectEl.value);
+        const response = await fetch(MISSION_PLAN_URL, {
+            method: "POST",
+            body: form,
+        });
+        if (!response.ok) {
+            let errorDetail = `Failed to set mission prompt (${response.status})`;
+            try {
+                const errJson = await response.json();
+                errorDetail = errJson.detail || errJson.error || errorDetail;
+            } catch (_) {}
+            throw new Error(errorDetail);
+        }
+        const payload = await response.json();
+        currentMissionId = payload.mission_id;
+        missionReady = true;
+        setStatus("Mission prompt ready. Upload videos to process.", "success");
+    } catch (error) {
+        console.error("[mission_plan] error", error);
+        setStatus(error.message || "Failed to set mission prompt.", "error");
+    } finally {
+        missionRequestPending = false;
+        updateControlState();
+    }
+}
 
 async function executeMission() {
 
-    const videoFile = document.getElementById("videoInput").files[0];
-    const mission = document.getElementById("missionPrompt").value;
-    const detector = document.getElementById("detectorSelect").value;
-    const summaryEl = document.getElementById("summary");
-    const statusEl = document.getElementById("status");
+    const videoFile = videoInputEl.files[0];
+    const mission = missionInputEl.value.trim();
+    const detector = detectorSelectEl.value;
 
-    if (!videoFile || !mission) {
-        alert("Mission invalid: Upload video and enter mission parameters.");
+    if (!currentMissionId || !missionReady) {
+        alert("Set the mission prompt before processing videos.");
+        return;
+    }
+    if (!videoFile) {
+        alert("Mission invalid: Upload a video file.");
         return;
     }
 
-    statusEl.textContent = "Processing video...";
+    setOriginalPreview(videoFile);
+
     summaryEl.textContent = "(Awaiting summary...)";
+    videoProcessing = true;
+    updateControlState();
+    setStatus("Processing video...");
 
     try {
         const videoForm = new FormData();
         videoForm.append("video", videoFile);
         videoForm.append("prompt", mission);
+        videoForm.append("mission_id", currentMissionId);
         videoForm.append("detector", detector);
 
         console.log("[process_video] submitting request", { detector, missionLength: mission.length, fileSize: videoFile.size });
@@ -207,11 +389,12 @@ async function executeMission() {
             console.error("[process_video] immediate video error", videoEl.error);
         }
 
-        statusEl.textContent = "Generating summary...";
+        setStatus("Generating summary...");
 
         const summaryForm = new FormData();
         summaryForm.append("video", videoFile);
         summaryForm.append("prompt", mission);
+        summaryForm.append("mission_id", currentMissionId);
         summaryForm.append("detector", detector);
 
         console.log("[mission_summary] submitting request", { detector, missionLength: mission.length, fileSize: videoFile.size });
@@ -237,12 +420,24 @@ async function executeMission() {
         const summaryJson = await summaryResponse.json();
         const summaryText = summaryJson.mission_summary || "No summary returned.";
         summaryEl.textContent = summaryText;
-        statusEl.textContent = "Mission complete.";
+        setStatus("Mission complete.", "success");
     } catch (err) {
         console.error(err);
         summaryEl.textContent = "Mission failed.";
-        statusEl.textContent = `Error: ${err.message}`;
+        setStatus(`Error: ${err.message}`, "error");
+    } finally {
+        videoProcessing = false;
+        updateControlState();
+    }
+}
+
+function setOriginalPreview(file) {
+    if (originalVideoUrl) {
+        URL.revokeObjectURL(originalVideoUrl);
     }
+    originalVideoUrl = URL.createObjectURL(file);
+    originalVideoEl.src = originalVideoUrl;
+    originalVideoEl.load();
 }
 </script>
 

inference.py

@@ -72,6 +72,7 @@ def run_inference(
     detector_name: Optional[str] = None,
     write_output_video: bool = True,
     generate_summary: bool = True,
+    mission_plan: Optional[MissionPlan] = None,
 ) -> Tuple[Optional[str], MissionPlan, Optional[str]]:
     try:
         frames, fps, width, height = extract_frames(input_video_path)
@@ -79,9 +80,12 @@ def run_inference(
         logging.exception("Failed to decode video at %s", input_video_path)
         raise
 
-    mission_plan = get_mission_plan(mission_prompt)
-    logging.info("Mission plan: %s", mission_plan.to_json())
-    queries = mission_plan.queries()
+    mission_prompt_clean = (mission_prompt or "").strip()
+    if not mission_prompt_clean:
+        raise ValueError("Mission prompt is required.")
+    resolved_plan = mission_plan or get_mission_plan(mission_prompt_clean)
+    logging.info("Mission plan: %s", resolved_plan.to_json())
+    queries = resolved_plan.queries()
 
     processed_frames: List[np.ndarray] = []
     detection_log: List[Dict[str, Any]] = []
@@ -101,6 +105,6 @@ def run_inference(
     else:
         video_path_result = None
     mission_summary = (
-        summarize_results(mission_prompt, mission_plan, detection_log) if generate_summary else None
+        summarize_results(mission_prompt_clean, resolved_plan, detection_log) if generate_summary else None
     )
-    return video_path_result, mission_plan, mission_summary
+    return video_path_result, resolved_plan, mission_summary