What: Research, Create, Tell, Iterate
Platform: Azure Container Apps + Foundry
Protocol: A2A (Agent-to-Agent)
App version: 1.12.0
Agent 1: Researcher
Checking...
LangGraph / Python
Actions & Tool Calls
- plan_research — Initialize research plan from user query
- search_docs — Search Microsoft Learn documentation
- search_code — Search GitHub Azure-Samples & Azure org repos
- search_blogs — Search Azure Blog, Tech Community & Azure Updates
- rank_sources — LLM-powered ranking of all discovered sources by relevance
- fetch_content — Fetch actual page content from top-ranked URLs
- synthesize — LLM synthesis of fetched content into a research brief
Agent 2: Content Creator
Checking...
Microsoft Agent Framework / .NET
Actions & Tool Calls
- BriefInputExecutor — Parse and validate the incoming research brief
- BlogGenerationExecutor — Generate a long-form blog article via LLM
- SocialGenerationExecutor — Generate LinkedIn post & Twitter thread via LLM
- OutputExecutor — Package blog + social results into content_package
Agent 3: Podcaster
Checking...
GitHub Copilot SDK / Python
Actions & Tool Calls
- fetch_content — Enrich source material by fetching original URLs
- generate_script — Generate a two-host conversation script via LLM
- dynamic_pronunciations — Generate & apply pronunciation fixes for TTS
- synthesize_audio — Synthesize speech for each script turn via TTS
- assemble_audio — Interleave audio segments & convert to MP3
- upload_blob — Upload final audio to Azure Blob Storage
Agent 4: Video
Checking...
Google VEO / Python
Actions & Tool Calls
- generate_video — Generate a short MP4 from a text prompt via Google VEO
- upload_video — Upload generated MP4 to Azure Blob Storage with a 30-day SAS URL
Video Prompt
{Technology-Topic-Input} is replaced with the topic above at run time
What content do you want me to create and for what audience?
Agent Configuration
URL
A2A Card
OTel Agent ID
A2A Token
Platform
A2A Messages
No A2A messages yet. Run the pipeline to see agent-to-agent communication.
Run History
Previous Generations
Open this tab to load previous generations.
Run Details
Select a run to view its details.
Paste one or more URLs (or a wall of text containing links). When you press Store,
a single valid URL is sent straight to the persistence API. Anything else is first passed to the
decoder which extracts URLs via regex and stores them in bulk.
Stored links
Open this tab to load stored links.
Jobs
Review background processor executions stored in Cosmos DB, inspect details, delete old entries, or trigger the Azure Container Apps Job.
Job Executions
Open this tab to load job executions.
Job Details
Select a job execution to view its details.
YouTube transcript
Capture a YouTube video as a transcript + LLM summary. YouTube no longer permits reliable third-party caption download, so paste the transcript text manually (open the video on YouTube → "..." → "Show transcript" → select all → copy). The agent persists it in Cosmos, generates a structured summary, and patches it onto the same document.
Status
Paste a URL and the transcript text, then submit.
Summary
The generated summary will appear here once the task completes.
Saved transcripts
Open this tab to load saved transcripts.