Implementation Guide: Capture After-Action Reviews & EOC Briefings — Generate SITREPs and Lessons-Learned Records

# verify ANIUSB-MATRIX is detected as an audio endpoint

Get-PnpDevice -Class AudioEndpoint | Where-Object {$_.FriendlyName -like '*ANIUSB*'}

# streams microphone audio to Azure Government Speech endpoint with real-
# time file output

# eoc_transcription_service.py
# Runs on EOC recording workstation — streams to Azure Government Speech endpoint

import azure.cognitiveservices.speech as speechsdk
import datetime
import os

SPEECH_KEY = os.environ.get("AZURE_SPEECH_KEY")
SPEECH_REGION = "usgovvirginia"  # Azure Government region

def create_transcript_file():
    timestamp = datetime.datetime.now().strftime("%Y%m%d_%H%M%S")
    return f"eoc_briefing_{timestamp}.txt"

def transcribe_eoc_briefing():
    speech_config = speechsdk.SpeechConfig(
        subscription=SPEECH_KEY,
        region=SPEECH_REGION
    )
    speech_config.speech_recognition_language = "en-US"
    speech_config.set_property(
        speechsdk.PropertyId.SpeechServiceConnection_EndSilenceTimeoutMs, "3000"
    )

    # Use default microphone input (ANIUSB-MATRIX set as Windows default)
    audio_config = speechsdk.audio.AudioConfig(use_default_microphone=True)

    # Conversation transcriber supports speaker identification
    transcriber = speechsdk.transcription.ConversationTranscriber(
        speech_config=speech_config,
        audio_config=audio_config
    )

    transcript_file = create_transcript_file()
    all_text = []

    def handle_transcribed(evt):
        if evt.result.reason == speechsdk.ResultReason.RecognizedSpeech:
            timestamp = datetime.datetime.now().strftime("%H:%M:%S")
            speaker = evt.result.speaker_id or "Unknown"
            line = f"[{timestamp}] [{speaker}]: {evt.result.text}"
            print(line)
            all_text.append(line)
            # Append to file in real time for resilience
            with open(transcript_file, 'a', encoding='utf-8') as f:
                f.write(line + "\n")

    def handle_canceled(evt):
        print(f"Transcription canceled: {evt.reason}")
        if evt.reason == speechsdk.CancellationReason.Error:
            print(f"Error: {evt.error_details}")

    transcriber.transcribed.connect(handle_transcribed)
    transcriber.canceled.connect(handle_canceled)

    print(f"EOC Briefing Transcription Active — output: {transcript_file}")
    print("Press Enter to stop transcription and generate SITREP...")

    transcriber.start_transcribing_async()
    input()  # Wait for operator to press Enter at briefing end
    transcriber.stop_transcribing_async()

    print(f"\nTranscription complete. {len(all_text)} segments captured.")
    return transcript_file, "\n".join(all_text)

if __name__ == "__main__":
    transcript_file, full_transcript = transcribe_eoc_briefing()
    print(f"Transcript saved to: {transcript_file}")
    # Pass to SITREP generation function

# Generates ICS-compliant SITREP from EOC briefing transcript

# sitrep_generator.py
# Generates ICS-compliant SITREP from EOC briefing transcript

from openai import AzureOpenAI
import os
import datetime

client = AzureOpenAI(
    azure_endpoint=os.environ["AZURE_OPENAI_ENDPOINT"],  # *.openai.azure.us
    api_key=os.environ["AZURE_OPENAI_KEY"],
    api_version="2024-08-01-preview"
)

SITREP_PROMPT = """
You are an emergency management documentation specialist trained in FEMA NIMS/ICS protocols.
Generate a Situation Report (SITREP) from the following EOC briefing transcript.

Format the SITREP using the following standard structure:

---
SITUATION REPORT
Incident Name: [Extract from transcript or 'TBD']
SITREP Number: [Sequence — use 1 if not determinable]
Reporting Period: [Date/Time from transcript or today's date]
Prepared by: [Documentation Unit — to be completed by reviewer]
Approved by: [Incident Commander — LEAVE BLANK for human signature]
Classification: UNCLASSIFIED [flag if any sensitive content detected]

1. SITUATION OVERVIEW
   [2-3 sentence summary of current incident status]

2. INCIDENT OBJECTIVES (from briefing)
   [List objectives discussed — numbered]

3. CURRENT STATUS BY SECTION
   Operations: [Summary of operational status]
   Logistics: [Summary of logistics/resource status]
   Planning: [Summary of planning activities]
   Finance/Admin: [Summary if discussed]
   Public Information: [Summary if discussed]

4. RESOURCE SUMMARY
   Resources Deployed: [List if mentioned]
   Resources Requested/Pending: [List if mentioned]
   Resource Shortfalls: [List if mentioned]

5. PRIORITIES FOR NEXT OPERATIONAL PERIOD
   [Numbered list of priorities discussed]

6. SIGNIFICANT EVENTS (past 12-24 hours)
   [Bulleted list of significant events mentioned]

7. ANTICIPATED ACTIONS (next 12-24 hours)
   [Bulleted list of anticipated activities]

8. ISSUES AND CONCERNS
   [List unresolved issues or concerns raised in briefing]

9. WEATHER / ENVIRONMENTAL CONDITIONS
   [If discussed in briefing]

10. PUBLIC INFORMATION SUMMARY
    [Key public messaging points if discussed]

REVIEWER INSTRUCTIONS:
- Items marked [VERIFY] require confirmation before distribution
- Items marked [SENSITIVE] should be reviewed for appropriate handling before external distribution
- All AI-generated content must be reviewed and approved by Documentation Unit Leader before distribution
- Do not distribute this SITREP until approved by Incident Commander or designee

---

BRIEFING TRANSCRIPT:
{transcript}
"""

def generate_sitrep(transcript_text: str, incident_name: str = "Unknown") -> str:
    response = client.chat.completions.create(
        model=os.environ["AZURE_OPENAI_DEPLOYMENT"],
        messages=[
            {
                "role": "system",
                "content": "You are an emergency management documentation specialist. Generate accurate, concise, and actionable emergency management documents. Flag any content that appears sensitive or requires human verification."
            },
            {
                "role": "user",
                "content": SITREP_PROMPT.format(transcript=transcript_text)
            }
        ],
        temperature=0.1,
        max_tokens=3000
    )

    sitrep = response.choices[0].message.content
    timestamp = datetime.datetime.now().strftime("%Y%m%d_%H%M%S")
    filename = f"SITREP_DRAFT_{incident_name}_{timestamp}.md"

    with open(filename, 'w', encoding='utf-8') as f:
        f.write(f"[DRAFT — AI GENERATED — REQUIRES HUMAN REVIEW AND APPROVAL]\n\n")
        f.write(sitrep)

    print(f"Draft SITREP saved: {filename}")
    return filename

if __name__ == "__main__":
    # Read transcript from file
    import sys
    if len(sys.argv) > 1:
        with open(sys.argv[1], 'r') as f:
            transcript = f.read()
        generate_sitrep(transcript, incident_name=sys.argv[2] if len(sys.argv) > 2 else "Incident")

# Generates a structured After-Action Report from a training/exercise
# transcript using Azure OpenAI

# aar_generator.py
# Generates structured After-Action Report from training/exercise transcript

AAR_PROMPT = """
You are a military training and readiness specialist. Generate a structured 
After-Action Report (AAR) from the following session transcript.

Follow the standard AAR format per Army Training Circular 25-20:

---
AFTER-ACTION REPORT
Unit/Organization: [Extract or 'TBD']
Exercise/Event Name: [Extract or 'TBD']
Date of Event: [Extract or today's date]
Location: [Extract or 'TBD']
Prepared by: [AAR Facilitator — to be completed by reviewer]
Classification: UNCLASSIFIED [Verify and adjust if needed]

1. EXECUTIVE SUMMARY
   [2-3 sentence overview of the event and key findings]

2. TASK(S) EVALUATED
   [List the tasks, missions, or objectives that were the focus of the event]

3. WHAT WAS SUPPOSED TO HAPPEN (Standard/Plan)
   [Expected outcomes, planned actions, and standards against which performance is measured]

4. WHAT ACTUALLY HAPPENED (Observations)
   [Factual account of what occurred — no judgment, just facts from the discussion]

5. WHY IT HAPPENED (Discussion)
   For each significant observation:
   - Observation: [What was observed]
   - Root Cause: [Why it happened — contributing factors]
   - Impact: [Effect on mission/training objective]

6. STRENGTHS (Sustain)
   [Numbered list of things that went well and should be repeated/reinforced]

7. AREAS FOR IMPROVEMENT (Improve)
   [Numbered list of things that need to be improved]

8. LESSONS LEARNED
   [Transferable lessons applicable beyond this specific event]

9. RECOMMENDED ACTIONS
   For each area for improvement:
   - Action: [Specific corrective action]
   - Who: [Responsible party]
   - By When: [Timeline]
   - How Measured: [How completion/success is determined]

10. CONCLUSION
    [Brief summary of overall assessment and path forward]

REVIEWER INSTRUCTIONS:
- Mark with [CLASSIFICATION REVIEW NEEDED] if any discussion touched on sensitive topics
- Verify all recommended actions are specific and achievable
- Have the senior leader present sign/approve before distribution

---
TRANSCRIPT:
{transcript}
"""

def generate_aar(transcript_text: str, event_name: str = "Training Event") -> str:
    from openai import AzureOpenAI
    import os, datetime

    client = AzureOpenAI(
        azure_endpoint=os.environ["AZURE_OPENAI_ENDPOINT"],
        api_key=os.environ["AZURE_OPENAI_KEY"],
        api_version="2024-08-01-preview"
    )

    response = client.chat.completions.create(
        model=os.environ["AZURE_OPENAI_DEPLOYMENT"],
        messages=[
            {"role": "system", "content": "You are a military training documentation specialist. Generate factual, structured after-action reports. Do not embellish or add observations not supported by the transcript."},
            {"role": "user", "content": AAR_PROMPT.format(transcript=transcript_text)}
        ],
        temperature=0.1,
        max_tokens=4000
    )

    aar = response.choices[0].message.content
    timestamp = datetime.datetime.now().strftime("%Y%m%d_%H%M%S")
    filename = f"AAR_DRAFT_{event_name.replace(' ','_')}_{timestamp}.md"

    with open(filename, 'w', encoding='utf-8') as f:
        f.write(f"[DRAFT — AI GENERATED — REQUIRES FACILITATOR REVIEW AND COMMANDER APPROVAL]\n\n")
        f.write(aar)

    return filename

# Pushes AI-generated SITREP to WebEOC via REST API

# webeoc_integration.py
# Pushes AI-generated SITREP to WebEOC via REST API

import requests
import os
import json

WEBEOC_BASE_URL = os.environ["WEBEOC_URL"]  # e.g., https://webeoc.agency.gov
WEBEOC_USERNAME = os.environ["WEBEOC_SERVICE_ACCOUNT"]
WEBEOC_PASSWORD = os.environ["WEBEOC_SERVICE_PASSWORD"]
WEBEOC_INCIDENT = os.environ["WEBEOC_INCIDENT_ID"]

def get_webeoc_token():
    """Authenticate to WebEOC and retrieve session token."""
    auth_url = f"{WEBEOC_BASE_URL}/api/authenticate"
    resp = requests.post(auth_url, json={
        "username": WEBEOC_USERNAME,
        "password": WEBEOC_PASSWORD
    }, verify=True)  # Always verify SSL in government environments
    resp.raise_for_status()
    return resp.json().get("token")

def create_sitrep_in_webeoc(sitrep_content: str, incident_id: str):
    """Create a draft SITREP record in WebEOC."""
    token = get_webeoc_token()
    headers = {
        "Authorization": f"Bearer {token}",
        "Content-Type": "application/json"
    }

    # WebEOC board/record structure varies by agency configuration
    # Adjust field names to match the agency's WebEOC SITREP board schema
    sitrep_record = {
        "incidentId": incident_id,
        "boardName": "Situation Reports",
        "status": "DRAFT",
        "preparedBy": "AI Documentation Assistant — REQUIRES HUMAN REVIEW",
        "content": sitrep_content,
        "distribution": "INTERNAL ONLY — NOT FOR EXTERNAL DISTRIBUTION UNTIL APPROVED",
        "autoGenerated": True,
        "requiresApproval": True
    }

    create_url = f"{WEBEOC_BASE_URL}/api/incidents/{incident_id}/boards/sitreps"
    resp = requests.post(create_url, json=sitrep_record, headers=headers, verify=True)
    resp.raise_for_status()

    record_id = resp.json().get("id")
    print(f"Draft SITREP created in WebEOC. Record ID: {record_id}")
    print(f"URL: {WEBEOC_BASE_URL}/incidents/{incident_id}/sitreps/{record_id}")
    return record_id

SYSTEM PROMPT:
You are an exercise evaluation specialist certified in FEMA's Homeland Security 
Exercise and Evaluation Program (HSEEP). Based on the provided After-Action Report,
generate a structured Improvement Plan (IP) in HSEEP format.

For each Area for Improvement (AFI) identified in the AAR, produce:

IMPROVEMENT PLAN
Exercise Name: [From AAR]
Date: [From AAR]
Prepared by: [To be completed by reviewer]

IMPROVEMENT MATRIX:
| # | Capability | AFI Description | Recommendation | Responsible Agency | Point of Contact | Due Date | Status |
|---|---|---|---|---|---|---|---|
[Populate from AAR findings]

IMPLEMENTATION NOTES:
- Each recommendation must be SMART (Specific, Measurable, Achievable, Relevant, Time-bound)
- Due dates should be realistic — default to 90 days unless context suggests otherwise
- Flag recommendations requiring interagency coordination with [MULTI-AGENCY]
- Flag recommendations requiring funding with [FUNDING REQUIRED]
- This IP should be reviewed at the post-exercise conference and updated quarterly

INPUT AAR:
{aar_content}

SYSTEM PROMPT:
You are an emergency management quality assurance reviewer.
Review the following draft SITREP and evaluate it against these criteria:

COMPLETENESS CHECK:
- [ ] Incident name and number present
- [ ] Reporting period clearly stated
- [ ] All standard ICS sections addressed (or noted as N/A)
- [ ] Resource summary present
- [ ] Next operational period priorities identified
- [ ] No blank or 'TBD' fields that should have been populated from the transcript

ACCURACY FLAGS:
- Flag any numeric claims (resource counts, casualty figures, dates) with [VERIFY-NUMBER]
- Flag any named individuals with [VERIFY-NAME]
- Flag any claims not clearly supported by the transcript with [UNVERIFIED]

SENSITIVE CONTENT FLAGS:
- [CIKR] — any mention of critical infrastructure specifics
- [VICTIM-PHI] — any individually identifiable victim information
- [LAW-ENFORCEMENT-SENSITIVE] — any tactical law enforcement information
- [OPSEC] — any information that could compromise ongoing response operations

DISTRIBUTION READINESS:
Rate each distribution channel:
- Internal EOC only: [READY / NOT READY — reason]
- State EOC: [READY / NOT READY — reason]
- Public (PIO): [READY / NOT READY — reason]
- HSIN: [READY / NOT READY — reason]

DRAFT SITREP:
{sitrep_content}