> ## Documentation Index
> Fetch the complete documentation index at: https://praison.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# Platform Workspace Context
> Provide workspace-level context and agent configuration from the platform database
Workspace context lets platform-hosted agents load settings and instructions scoped to a workspace.
```python theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
from praisonaiagents import Agent
from praisonaiagents.auth import WorkspaceContextProtocol
class PlatformWorkspaceContext:
async def get_workspace_context(self, workspace_id: str) -> str | None:
# Load workspace instructions from your database
return "You support the Acme Corp workspace."
async def get_agent_config(self, workspace_id: str, agent_id: str) -> dict | None:
return {
"system_prompt": "Be concise and professional.",
"model": "gpt-4o-mini",
"tools": ["web_search"],
"max_concurrent_tasks": 3,
}
agent = Agent(
name="workspace-assistant",
instructions="Help users in their workspace.",
)
```
The user opens a workspace thread; the agent loads scoped instructions from platform context.
```mermaid theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
graph LR
Agent[🤖 Agent] --> Ctx[📋 WorkspaceContext]
Ctx --> WS[🏢 Workspace Settings]
Ctx --> Config[⚙️ Agent Config]
WS --> Prompt[📝 System Prompt]
Config --> Prompt
classDef agent fill:#8B0000,stroke:#7C90A0,color:#fff
classDef ctx fill:#189AB4,stroke:#7C90A0,color:#fff
classDef output fill:#10B981,stroke:#7C90A0,color:#fff
class Agent agent
class Ctx,WS,Config ctx
class Prompt output
```
## Quick Start
Implement `WorkspaceContextProtocol` with workspace-level instructions:
```python theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
from praisonaiagents.auth import WorkspaceContextProtocol
class PlatformWorkspaceContext:
def __init__(self, db):
self.db = db
async def get_workspace_context(self, workspace_id: str) -> str | None:
workspace = await self.db.get_workspace(workspace_id)
if not workspace:
return None
return workspace.instructions
async def get_agent_config(self, workspace_id: str, agent_id: str) -> dict | None:
agent = await self.db.get_agent(workspace_id, agent_id)
if not agent:
return None
return {
"system_prompt": agent.system_prompt,
"model": agent.model,
"tools": agent.tools,
"max_concurrent_tasks": agent.max_concurrent_tasks,
}
```
Merge workspace and agent settings before starting the agent:
```python theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
from praisonaiagents import Agent
async def start_platform_agent(ctx: PlatformWorkspaceContext, workspace_id: str, agent_id: str):
workspace_text = await ctx.get_workspace_context(workspace_id)
config = await ctx.get_agent_config(workspace_id, agent_id) or {}
instructions = config.get("system_prompt", "You are a helpful assistant.")
if workspace_text:
instructions = f"{workspace_text}\n\n{instructions}"
agent = Agent(
name=config.get("name", "assistant"),
instructions=instructions,
llm=config.get("model"),
)
return agent
```
***
## How It Works
```mermaid theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
sequenceDiagram
participant Platform
participant Ctx as WorkspaceContext
participant DB
participant Agent
Platform->>Ctx: get_workspace_context(ws_id)
Ctx->>DB: fetch workspace
DB-->>Ctx: instructions
Platform->>Ctx: get_agent_config(ws_id, agent_id)
Ctx->>DB: fetch agent (scoped)
DB-->>Ctx: config dict
Ctx-->>Agent: merged prompt + settings
```
| Method | Returns |
| ------------------------------------------ | -------------------------------------------------------------------------- |
| `get_workspace_context(workspace_id)` | Workspace instructions string, or `None` |
| `get_agent_config(workspace_id, agent_id)` | Config dict with `system_prompt`, `model`, `tools`, `max_concurrent_tasks` |
***
## Configuration Options
| Protocol field | Type | Description |
| ----------------------- | ----------------------------------------------- | ---------------------------- |
| `get_workspace_context` | `async (workspace_id) → str \| None` | Workspace-level instructions |
| `get_agent_config` | `async (workspace_id, agent_id) → dict \| None` | Per-agent runtime settings |
Expected keys in the agent config dict: `system_prompt`, `model`, `tools`, `max_concurrent_tasks`.
***
## Best Practices
Filter by both `agent_id` and `workspace_id` — never fetch agents globally.
Missing workspace or agent records should return `None`, not raise.
Pass `AsyncSession` via constructor — do not create sessions inside the context provider.
Convert UUIDs and datetimes to strings before returning config dicts.
***
## Related
Authentication and authorisation setup
Register and manage platform-hosted agents