Meeting Extraction, Monitoring & Orchestration
● System Healthy
Total Agents
6
3 extraction · 3 decision
Tests Passing
17 / 17
100% pass rate
Action Items Extracted
17
From last run · 9 assigned
Recovery Success Rate
100%
2 recoveries · 0 escalated
Agent Confidence Scores
TranscriptAnalyzer
DecisionExtractor
ActionItemExtractor
TaskPrioritizer
OwnerAssigner
EscalationDecider
Meeting Participants & Activity
ParticipantRoleTasks AssignedStatus
Sarah ChenManager2Active
Bob WilsonDeveloper3High Load
Maria GarciaEng. Lead3High Load
John SmithMarketing Lead2Active
Top Extracted Action Items
IDTaskAssigned ToDeadlinePriorityStatus
AI-001Complete security audit documentationMaria GarciaMondayP1In Progress
AI-002Resolve API integration with vendorBob WilsonWednesday EODP1Blocked
AI-003Update engineering scheduleMaria GarciaFridayP2On Track
AI-004Review new landing page contentSarah ChenNext weekP2On Track
AI-005Coordinate DevOps CI/CD investigationJohn SmithTodayP2Started
AI-006Update campaign materials (new launch date)MarketingJan 10P3Planned
AI-007Schedule follow-up progress meetingSarah ChenNext TuesdayP3Planned

📝 What is MEMO?

MEMO (Meeting Extraction, Monitoring & Orchestration) is a multi-agent intelligence system that transforms raw, unstructured meeting transcripts into structured, actionable intelligence. Instead of manually reviewing meeting notes, MEMO automatically identifies who said what, what decisions were made, what action items were assigned, and which items need immediate escalation.

⚡ How It Works

You paste or upload a meeting transcript. MEMO processes it through a pipeline of 6 autonomous agents that run sequentially, each building on the previous agent's output:

Step 1 — Extraction
🔍 TranscriptAnalyzer
Parses the raw transcript to identify speakers, topics, and conversation segments. Structures the unformatted text into analyzable sections.
Step 2 — Extraction
📋 DecisionExtractor
Identifies explicit and implicit decisions made during the meeting. Captures decision context, rationale, and who was involved in making each decision.
Step 3 — Extraction
✅ ActionItemExtractor
Pulls out all action items, to-dos, and follow-ups mentioned during the meeting. Captures deadlines, responsible parties, and dependencies.
Step 4 — Decision
📊 TaskPrioritizer
Evaluates each extracted action item and assigns a priority level (Critical, High, Medium, Low) based on urgency, impact, and deadlines.
Step 5 — Decision
👤 OwnerAssigner
Determines who should own each action item based on context clues, mentions, and responsibility patterns identified in the transcript.
Step 6 — Decision
🚨 EscalationDecider
Evaluates whether any items require immediate escalation — blocked tasks, missed deadlines, cross-team dependencies, or unresolved conflicts.

🎬 Live Agent Pipeline Demo

Watch how MEMO processes a meeting transcript through all 6 agents in real-time

Alice: Let's discuss the Q3 roadmap. The API deadline is Friday.
Bob: I'll handle the backend integration. Need auth tokens from DevOps.
Alice: Agreed. Bob, please have the API docs ready by Thursday.
Charlie: I'm blocked on the frontend — waiting for Bob's endpoints.
🔎
Step 1
TranscriptAnalyzer
Idle
3 speakers · 4 segments parsed
📋
Step 2
DecisionExtractor
Idle
2 decisions: API priority + deadline
Step 3
ActionItemExtractor
Idle
3 actions: API, docs, auth tokens
📈
Step 4
TaskPrioritizer
Idle
P1: API · P2: Docs, Tokens
👤
Step 5
OwnerAssigner
Idle
Bob → API, Docs · DevOps → Tokens
🚨
Step 6
EscalationDecider
Idle
⚠ Charlie blocked → Escalate
✅ Pipeline Complete — 6 agents processed in 4.2s
2
Decisions
3
Action Items
3
Assigned
1
Escalations

🎯 Real-World Use Cases

Engineering Standups
Automatically extract blockers, decisions, and assigned tasks from daily standup transcripts — no more forgotten action items.
Board Meeting Minutes
Turn lengthy board meeting recordings into structured minutes with clear decisions, action items, and accountability assignments.
Client Call Tracking
Capture client requirements, commitments, and follow-up tasks from sales or customer success calls to ensure nothing falls through the cracks.
Sprint Planning & Retrospectives
Extract sprint goals, assigned stories, retrospective action items, and improvement commitments from agile ceremonies.

🏗️ Architecture

🔄
Orchestration Engine
Coordinates all 6 agents in sequence, managing state and message passing between each step.
🛡️
Circuit Breaker Recovery
Each agent has its own circuit breaker. If failures spike, the circuit opens to prevent cascading failures.
📋
Full Audit Trail
Every decision, confidence score, and state change is logged immutably for compliance and traceability.
📤 Upload Your Own Transcript LIVE AGENTS
📄
Drag & drop a file here, or click to browse
.txt · .md · .log supported
✦ Your transcript is sent to the live Python agent pipeline — real extraction, real decisions, auditable results.
Live Agent Results — Your Transcript
Agent Breakdown
Extracted Action Items
TaskAssigned ToUrgencyDeadline
Extracted Decisions
TypeDecision

— or run a sample transcript —
📝
Transcript
Analyzer
Extraction
Idle
🔍
Decision
Extractor
Extraction
Idle
Action Item
Extractor
Extraction
Idle
📊
Task
Prioritizer
Decision
Idle
👤
Owner
Assigner
Decision
Idle
🚨
Escalation
Decider
Decision
Idle
Execution Log
Click "Run" to start the pipeline simulation…
Decisions Found
Explicit + assignments
Action Items
Tasks extracted
Exec Time
End-to-end
Source Transcript — Q4 Product Launch Planning
Total Tests
17
Passed
17
Failed
0
Duration
2.32s
Test Suites
TestTranscriptAnalyzer
test_basic_analysis0.21sPASSED
test_empty_transcript0.09sPASSED
TestDecisionExtractor
test_decision_extraction0.18sPASSED
TestActionItemExtractor
test_action_extraction0.31sPASSED
TestTaskPrioritizer
test_prioritization0.14sPASSED
TestOwnerAssigner
test_owner_assignment0.17sPASSED
Test Suites (continued)
TestRecoveryStrategies
test_retry_strategy0.05sPASSED
test_recovery_manager0.06sPASSED
TestCircuitBreaker
test_circuit_closed_allows_calls0.04sPASSED
test_circuit_opens_on_failures0.03sPASSED
TestAuditLogger
test_log_event0.08sPASSED
test_audit_trail_export0.12sPASSED
TestStateManager
test_workflow_state_lifecycle0.09sPASSED
test_shared_data0.07sPASSED
test_snapshots0.11sPASSED
TestMessageBus
test_send_receive0.05sPASSED
test_broadcast0.04sPASSED
Raw pytest Output
08:31:00RUNpython -m pytest tests/test_agents.py -v --tb=short --no-header
08:31:00COLLECTcollected 17 items
08:31:00PASSTestTranscriptAnalyzer::test_basic_analysis [ 5%]
08:31:00PASSTestTranscriptAnalyzer::test_empty_transcript [11%]
08:31:01PASSTestDecisionExtractor::test_decision_extraction [17%]
08:31:01PASSTestActionItemExtractor::test_action_extraction [23%]
08:31:01PASSTestTaskPrioritizer::test_prioritization [29%]
08:31:01PASSTestOwnerAssigner::test_owner_assignment [35%]
08:31:01PASSTestRecoveryStrategies::test_retry_strategy [41%]
08:31:01PASSTestRecoveryStrategies::test_recovery_manager [47%]
08:31:01PASSTestCircuitBreaker::test_circuit_closed_allows_calls [52%]
08:31:01PASSTestCircuitBreaker::test_circuit_opens_on_failures [58%]
08:31:01PASSTestAuditLogger::test_log_event [64%]
08:31:02PASSTestAuditLogger::test_audit_trail_export [70%]
08:31:02PASSTestStateManager::test_workflow_state_lifecycle [76%]
08:31:02PASSTestStateManager::test_shared_data [82%]
08:31:02PASSTestStateManager::test_snapshots [88%]
08:31:02PASSTestMessageBus::test_send_receive [94%]
08:31:02PASSTestMessageBus::test_broadcast [100%]
08:31:02RESULT17 passed, 97 warnings in 2.32s ✓
Total Events
24
Last workflow run
Decisions Logged
8
With full reasoning
Avg Confidence
79%
Across all agents
Escalations
1
Auto-escalated to lead
Decision Audit Timeline
08:26:32 · workflow
Workflow Started
Meeting Intelligence Processor initialised
08:26:32 · agent
TranscriptAnalyzer → Complete
Extracted 5 speakers, 1 topic, 4 key points
Confidence 90%
08:26:33 · decision
Decision: Launch Delayed to Jan 10
Explicit decision — Sarah Chen (manager)
Confidence 87%
08:26:33 · decision
Priority Assigned: Security Audit → P1
High business impact · deadline within 3 days
Confidence 85%
08:26:34 · decision
Owner Assigned: API Integration → Bob Wilson
Skills: api, integration · Lowest workload (1 task)
Confidence 85%
08:26:34 · escalation
Escalation Triggered: API Integration Blocked
Deadline approaching · external vendor dependency
⚠ Escalated to Team Lead
08:26:35 · workflow
Workflow Completed
All steps finished · audit trail exported
Detailed Decision Reasoning
🎯 task_assignment — Bob Wilson
• Task requires API integration skills
• Bob Wilson has matching skills: api, integration
• Bob Wilson has lowest current workload (1 task)
• Selected with confidence 0.85
Confidence: 85%
⚠ priority_assigned — Security Audit P1
• High business impact: security keyword detected
• Deadline within 3 days of current date
• Business priority match: "security"
• Scored 72/100 → P1 High
Confidence: 85%
📋 decision_extracted — Launch Date
• Keyword "decided to" detected → explicit decision
• Speaker: Sarah Chen (role: manager)
• Full agreement: all participants present
• New deadline: January 10th
Confidence: 87%
Total Recoveries
2
Success Rate
100%
Retries Fired
1
Human Escalations
0
Fully autonomous
Circuit Breaker Status
TranscriptAnalyzer
CLOSED
0 failures
DecisionExtractor
CLOSED
0 failures
ActionItemExtractor
CLOSED
0 failures
TaskPrioritizer
CLOSED
0 failures
OwnerAssigner
CLOSED
0 failures
EscalationDecider
CLOSED
0 failures
Recovery Strategies Available
StrategyTriggerActionUsed
🔄 RetryStrategyTemporary errorRetry with backoff
↩ FallbackStrategyRepeated failuresSimplified processing
⏭ SkipStrategyNon-critical stepSkip and continue
↙ RollbackStrategyState corruptionRestore snapshot
📢 EscalateStrategyAll else failsNotify human lead
Recovery Event Log
08:26:33DETECTStep "api_call" failed — ConnectionError (attempt 1/3)
08:26:33RECOVERRetryStrategy selected — waiting 1.0s before retry
08:26:34RETRYStep "api_call" succeeded on attempt 2 ✓
08:26:34DETECTStep "vendor_sync" failed 3 consecutive times — threshold reached
08:26:34CBCircuit breaker for "vendor_sync" → OPEN
08:26:34RECOVEREscalateStrategy selected — escalating to team_lead
08:26:34ESCALATEEscalation notification sent to team_lead with full context ✓
08:26:35RESULTRecovery complete — 2/2 scenarios handled autonomously