Skip to content
Draft
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
98 changes: 98 additions & 0 deletions AGENTS.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,98 @@
# AGENTS.md: Drasi Platform

## 1. High-Level Overview

- **Purpose**: Drasi is a platform for building real-time, change-driven systems using continuous queries.
- **Architecture**: Polyglot microservices system designed for Kubernetes.
- **Core Technologies**: Rust (backend), Go (CLI), Dapr (inter-service communication).
- **Repository**: `https://github.com/drasi-project/drasi-platform/`.
- **Contribution Workflow**: Fork repo, commit to your fork and submit pull request.

### Architectural Flow Diagram

This diagram shows two primary flows: management (solid lines) and data pipeline (dashed lines).

```mermaid
graph TD
subgraph "User Interaction"
User["User/Developer"] --> CLI["cli/"];
end

subgraph "Platform Management"
CLI -->|Manages Resources via API| ControlPlane["control-planes/"];
ControlPlane -->|Deploys & Configures| DeployedServices;
end

subgraph "Deployed Services (on Kubernetes)"
direction LR
DeployedServices("Sources, Query Containers, Reactions");
end

subgraph "Core Data Pipeline"
direction LR
ExternalDataSource["External Data Source"] -.-> Sources["sources/"];
Sources -.-> QueryContainer["query-container/"];
QueryContainer -.-> Reactions["reactions/"];
Reactions -.-> ExternalSystem["External System (e.g., API, DB)"];
end
```

## 2. Critical Development Notes

### **Git Submodule: `drasi-core`**

The `query-container/query-host` service depends on the `drasi-core` git submodule. Before building the `query-container`, the submodule **must** be initialized. Other components (`reactions`, `sources`, `control-planes`) can be built without it.

- **Location**: `query-container/query-host/drasi-core/`
- **Source Repository**: `https://github.com/drasi-project/drasi-core`
- **Initialization Command**: `git submodule update --init --recursive`

Failure to initialize will cause compilation errors in `query-container`. For details on the engine, see the `README.md` within the submodule directory.

### **Data Contracts: `typespec/`**

The `typespec/` directory is the **single source of truth** for all inter-service data contracts.

- **DO NOT** edit generated model files elsewhere.
- All data model changes **must** be made in the source `.tsp` files.
- After editing `.tsp` files, run the full generation workflow:
1. `npm install && npm run build` (in `typespec/`).
2. Run `quicktype` commands (defined in component `Makefile`s) to generate language-specific code from the updated JSON Schema.

Failure to follow this workflow will cause data contract mismatches.

## 3. Component Directory Guide

- **`cli/`**: Go CLI (`drasi`) for platform installation and management.
- **`control-planes/`**: Rust backend. Contains the `mgmt_api` and the `kubernetes_provider`.
- **`query-container/`**: Rust microservices for the continuous query execution environment (`publish-api`, `query-host`, `view-svc`).
- **`reactions/`**: Rust microservices that act on query results. Contains built-in reactions and SDKs.
- **`sources/`**: Rust microservices that ingest data from external systems. Contains built-in sources and SDKs.
- **`typespec/`**: Source of truth for all data contracts.
- **`e2e-tests/`**: Jest/TypeScript end-to-end tests.

## 4. Build and Test

The root `Makefile` provides top-level targets for the primary development workflows:
- **`make docker-build`**: Builds all container images for the platform.
- **`make test`**: Runs the test suites for all components.
- **`make lint-check`**: Performs all code quality and lint checks.

For local Kubernetes workflows, `make kind-load` and `make k3d-load` are available to load built images into a cluster.

The build system uses a recursive `make` structure. The root `Makefile` delegates targets (e.g., `docker-build`, `test`) to Makefiles in component subdirectories (`control-planes/`, `query-container/`, etc.), which in turn delegate to their own subdirectories. This orchestrates a full repository build from the top down.

## 5. Core Architecture & Runtime

- **Pluggable Resource Providers**: The control plane is platform-agnostic. The `mgmt_api` communicates with a resource provider via a shared contract (`resource_provider_api`). The `kubernetes_provider` is the current implementation, translating Drasi resources into Kubernetes objects. This design allows for future platform providers.

- **Dapr Integration**: Dapr is a mandatory dependency for the backend, handling all inter-service communication (Pub/Sub) and resource lifecycle management (Dapr Actors). `drasi init` installs Dapr if needed.

- **Deployment Modes**:
- **Kubernetes**: The standard deployment target. `drasi init` installs into the current `kubectl` context.
- **Docker**: `drasi init --docker` provides a self-contained environment using a K3d (Kubernetes in Docker) cluster.

- **User Workflow**:
1. Deploy the platform using `drasi init`.
2. Define `Source`, `ContinuousQuery`, and `Reaction` resources in YAML.
3. Deploy and manage resources using `drasi apply -f <file>.yaml`.
1 change: 1 addition & 0 deletions CLAUDE.md
81 changes: 81 additions & 0 deletions reactions/AGENTS.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,81 @@
# AGENTS.md: Drasi Reactions

## 1. Architectural Role

- **Position in Pipeline**: Reactions are the final stage in the Drasi data processing pipeline: **Sources -> Continuous Queries -> Reactions**.
- **Core Function**: They are event-driven microservices that subscribe to the real-time stream of changes (add, update, delete) from Continuous Queries and execute actions on external systems.
- **Responsibilities**:
- **Integration**: Act as a bridge between Drasi and external systems (e.g., message brokers, databases, APIs).
- **Action**: Execute business logic based on query result changes.
- **Decoupling**: Separate the change detection logic (the "what" in a query) from the action logic (the "how" in a reaction).

## 2. Core Abstraction: The `ReactionProvider` Model

The architecture separates a reaction's *type definition* from its *instance configuration* using two resource types: `ReactionProvider` and `Reaction`.

### `ReactionProvider`
- **Role**: Type Definition / Template.
- **Purpose**: A cluster-level resource that registers a **type** of reaction with the Drasi control plane. It serves as a template and validation schema for all reactions of its kind.
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

What do we mean by cluster-level?

- **Defines**:
- Container image for the reaction's implementation.
- Default Dapr settings and service configurations.
- A **`config_schema` (JSON Schema)** which serves as a validation contract for all `Reaction` instances of this type. The `mgmt_api` enforces this schema during creation and updates.
- **Lifecycle**: Registered once per reaction type. Built-in providers are registered during `drasi init`.

### `Reaction`
- **Role**: Instance / Configured Deployment.
- **Purpose**: A specific, configured instance of a `ReactionProvider`. This is the running microservice pod.
- **Defines**:
- A reference (`kind`) to its `ReactionProvider`.
- Specific `properties` (e.g., connection strings) that conform to the provider's `config_schema`.
- The list of `queries` it subscribes to.
- **Lifecycle**: Created, updated, and deleted as needed by users via the `drasi` CLI.

- **Key Takeaway**: This model enables a single containerized reaction implementation to be deployed multiple times, each with a unique configuration, subscription list, and lifecycle. For fully annotated YAML examples of both `ReactionProvider` and `Reaction` resources, refer to the `README.md` file in this directory.

## 3. Deployment Orchestration Flow

Applying a `Reaction` manifest via `drasi apply` initiates the following platform-agnostic orchestration flow:

1. **CLI to Management API**: The `drasi` CLI sends the `Reaction` manifest to the `/v1/reactions/{reaction-name}` endpoint.
2. **Validation & Persistence**: The Management API validates the `spec.properties` against the `config_schema` from the corresponding `ReactionProvider` and persists the desired state.
3. **Delegation to Resource Provider**: The Management API delegates the provisioning task to the active resource provider (e.g., `kubernetes_provider`) using the platform-agnostic `resource_provider_api` contract.
4. **Platform-Specific Resource Generation**: The active resource provider translates the abstract `Reaction` spec into concrete resources. For the default `kubernetes_provider`, this includes:
- Kubernetes `Deployment`
- Kubernetes `ConfigMap` (for query configurations)
- Kubernetes `Service`
- Dapr `Component` (for `pubsub.redis` subscription)
5. **Apply to Platform**: The resource provider applies the generated manifests to the target platform's API (e.g., Kubernetes API).

## 4. Configuration at Runtime

A running reaction container accesses its configuration via two mechanisms:

- **Environment Variables**: All key-value pairs from the `spec.properties` section of the `Reaction` manifest are injected as environment variables. Values from Kubernetes Secrets are securely resolved and injected.
- **Volume-Mounted Files**: The `spec.queries` map is stored in a `ConfigMap` and mounted as a file volume at `/etc/queries`.

## 5. Reaction Runtime Data Flow

This diagram illustrates the data flow *inside* a running Reaction pod.

```mermaid
graph TD
A[Continuous Query] -->|Change Results| B[Dapr Pub/Sub]
B -->|Event Stream| C[Reaction Pod]
C -->|Deserialize| D[Change Events]
D -->|Apply Query Config| E[Event Handler]
E -->|Execute Action| F[External System]

G[ConfigMap] -->|Query Configuration| C
H[Secrets] -->|Credentials| C
```

## 6. Examples of Reactions

- **`PostDaprPubSub`**: Forwards query results to a Dapr Pub/Sub topic for decoupled microservice architectures.
- **`SyncDaprStateStore`**: Materializes query results into a Dapr state store, creating a continuously updated data cache.
- **`SignalR`**: Exposes a SignalR endpoint to broadcast query changes for real-time UIs.

## 7. Reaction SDKs

Drasi provides SDKs for multiple languages (located in `reactions/sdk/`) for the development of custom reactions.
1 change: 1 addition & 0 deletions reactions/CLAUDE.md
Loading
Loading