Ayush Formulation Verifier with AutoGen Multi-Agent Systems and Resemble AI Voice Interface
Validating traditional Ayush medical formulations against regulatory pharmacopeia requires rigorous verification and accessible citizen interfaces. Build an AutoGen multi-agent system integrated with Resemble AI voice synthesis to parse formulation ingredients, verify compliance against official databases, and voice response summaries.
What you are building
The core problem, expected build, and operating context for this challenge.
Develop an AutoGen collaborative agent system where a compliance agent and research agent cross-examine traditional medicine compositions with audio output from Resemble AI.
How work is evaluated
Evaluates AutoGen agents on verifying plant ingredient safety and producing synthetic voice responses.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
toxic_ingredient_detection
Verifies that toxic/unpurified ingredients are correctly flagged.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
verification_precision
Precision of compliance status identification. • target: 0.98 • range: 0-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Implement multi-agent conversations using AutoGen AssistantAgent and UserProxyAgent.
Integrate Resemble AI TTS API for generating voice guidance in regional compliance workflows.
Establish validation rules against standardized Ayush herbal medicine pharmacopeias.
Reference links and supporting material
Herbal formulation entries containing botanical Latin names, traditional usages, and regulatory safety limits.
Challenge setup
Evaluates AutoGen agents on verifying plant ingredient safety and producing synthetic voice responses.
Challenge input
JSON with formulation name, botanical ingredient list, and recommended dosage.
AutoGen
Multi-agent conversational structure for compliance cross-examination
Resemble AI
Voice generation for multi-lingual medical guidance
Evaluated output
JSON with safety_status, flag_reasons, and audio_stream_url.
- Verifies that toxic/unpurified ingredients are correctly flagged.
- Precision of compliance status identification. • target: 0.98 • range: 0-1
- Benchmark: Ayush AI Safety Benchmark
- Verification Precision target: 0.98
- 1 public reference case
- Python execution harness
- Python sandbox (unavailable on Versalist)
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
- Resemble AI · Required
- AutoGen · Optional
- n8n · Optional
Evaluation contract
- toxic_ingredient_detection · Weight 1
- verification_precision · Weight 1
Metrics
- verification_precision · Target 0.98
Benchmarks
- Ayush AI Safety Benchmark
- biotech
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON with formulation name, botanical ingredient list, and recommended dosage.
Output: JSON with safety_status, flag_reasons, and audio_stream_url.
Execution harness: Python
Evaluation tools
- AutoGen · agent_framework
- Resemble AI · assistants_and_interfaces
Recipe freeze marker
No recipe freeze time is recorded.