# Spatial9 production solutions

Deployable code that runs governed catalog renders from the data and AI platforms enterprises already run, or from an all open source stack. Every solution follows the same rules: key in the platform's secret store, at most 3 renders at once, an `idempotency_key` per track and version, and results written back for audit.

| Platform | Folder | Pattern |
|---|---|---|
| Databricks | [databricks](databricks/README.md) | Unity Catalog connection, Lakeflow Job, Volumes, optional AI Gateway MCP Services |
| Databricks Solutions Accelerator | [databricks/solution-accelerator](databricks/solution-accelerator/README.md) | Asset Bundle with Auto Loader, MCP, A2A and REST API routes, audit table and gold views |
| Snowflake | [snowflake](snowflake/README.md) | External access integration, Python procedure, Task, Cortex Agent custom tool |
| Google Cloud and Vertex AI | [vertex-ai](vertex-ai/README.md) | ADK agent with `McpToolset`, Agent Engine, Cloud Run job and Scheduler, BigQuery |
| AWS and Amazon Bedrock | [bedrock](bedrock/README.md) | Strands agent, Lambda, Step Functions, S3, AgentCore Gateway |
| Open source stack | [open-source](open-source/README.md) | MinIO, Kafka, PostgreSQL, FFmpeg, FastAPI and Dagster with real-time A2A webhooks |
| Any Python runtime | [shared/spatial9_mcp.py](shared/spatial9_mcp.py) | The MCP handshake with `requests` only |
| Any Python runtime | [shared/spatial9_a2a.py](shared/spatial9_a2a.py) | Agent2Agent tasks: send, follow, cancel, webhooks |

Other orchestrators such as Prefect, Salesforce Agentforce, ServiceNow and IBM watsonx Orchestrate can reuse `shared/spatial9_mcp.py` or connect to `https://spatial9.ai/mcp` as a remote MCP server where they support one.

Microsoft Fabric is not supported.

Guides:

- Enterprise platforms: https://spatial9.ai/blog/spatial9-enterprise-agents-databricks-snowflake-vertex-bedrock
- Open source pipeline: https://spatial9.ai/blog/open-source-spatial-audio-pipeline-real-time-spatial9