Workspace-specific context for the Granger autonomous research ecosystem - a graph-reinforced network for research, verification, and fine-tuning with hub-and-spokes architecture.
Provides workspace-specific context for the **Granger Ecosystem** - a graph-reinforced autonomous network for research, verification, and fine-tuning. Use this skill when working on any Granger-related projects or when context about the ecosystem is needed.
This skill provides comprehensive context about:
Use this skill when:
The Granger ecosystem follows a **hub-and-spokes architecture**:
```
Granger Ecosystem
├── Hub (Communication & Orchestration)
├── Reinforcement Learning (Core Intelligence)
├── World Model (Self-Understanding & Prediction)
├── Test Reporting (Quality Assurance)
├── Spokes (Specialized Processing Modules)
└── User Interfaces (Human Interaction)
```
**Core Pipeline Flow:**
```
SPARTA → Marker → ArangoDB → Unsloth
↓
LLM Call (LLM access)
↓
Module Communicator (orchestration)
```
**Core Infrastructure (6 projects):**
**Processing Spokes (7 projects):**
**User Interfaces (3 projects):**
**MCP Services (3 projects):**
**CRITICAL: All Granger modules MUST follow GRANGER_MODULE_STANDARDS.md**
**Locked Dependencies:**
```toml
requires-python = ">=3.10.11"
dependencies = [
"numpy==1.26.4", # LOCKED - Do not change
"pandas>=2.2.3,<2.3.0", # Compatible with numpy
"pyarrow>=4.0.0,<20", # mlflow constraint
"pillow>=10.1.0,<11.0.0", # Security constraint
]
```
**Required Environment Variables:**
```bash
PYTHONPATH=./src # MUST be first line
GRANGER_HUB_URL=http://localhost:8000
ARANGODB_URL=http://localhost:8529
LLM_CALL_URL=http://localhost:8001
TEST_REPORTER_URL=http://localhost:8002
MODULE_NAME=project_name
MODULE_VERSION=1.0.0
ENABLE_RL_OPTIMIZATION=true
```
All Granger projects must support:
**RL-Enabled Projects:**
**Standard RL Pattern:**
```python
from rl_commons import ContextualBandit, OptimizationAgent
class ModuleOptimizer:
def __init__(self):
self.agent = ContextualBandit(
actions=["option_a", "option_b", "option_c"],
context_features=["feature_1", "feature_2"],
exploration_rate=0.1
)
def optimize_decision(self, context: dict) -> str:
return self.agent.select_action(context)
def report_outcome(self, action: str, reward: float):
self.agent.update(action, reward)
```
**Test Reporter Integration:**
```python
from claude_test_reporter import GrangerTestReporter
def run_tests():
reporter = GrangerTestReporter(
module_name="project_name",
test_suite="integration"
)
results = pytest.main()
reporter.submit_results(results)
```
**Requirements:**
**Before Starting Work:**
1. Check Module Communicator for pending coordination requests
2. Review RL optimization suggestions for your module
3. Ensure Test Reporter shows no critical failures for dependencies
**After Completing Work:**
1. Run full test suite and report to Test Reporter
2. Update module schema if APIs changed
3. Notify Module Communicator of completion status
Available commands when working in Granger workspace:
```bash
/granger:health # Check ecosystem health
/granger:integration-test # Run cross-module tests
/granger:report # Generate ecosystem-wide report
/granger:optimize # Optimize module interactions
```
**Required Documentation:**
**Example 1: Starting work on a new Granger module**
```bash
cd /home/graham/workspace/shared_claude_docs
cat docs/07_style_conventions/GRANGER_MODULE_STANDARDS.md
/granger:health
cd /home/graham/workspace/experiments/new_module
```
**Example 2: Integrating with Hub**
```python
from granger_hub import ModuleCommunicator
communicator = ModuleCommunicator(
module_name="my_module",
hub_url=os.getenv("GRANGER_HUB_URL")
)
communicator.register_schema(my_schema)
communicator.report_progress("processing", 0.5)
```
**Example 3: Quick navigation between projects**
```bash
cd /home/graham/workspace/experiments/granger_hub # Hub
cd /home/graham/workspace/experiments/rl_commons # RL
cd /home/graham/workspace/experiments/sparta # Step 1
cd /home/graham/workspace/experiments/marker # Step 2
cd /home/graham/workspace/experiments/arangodb # Step 3
cd /home/graham/workspace/experiments/llm_call # LLM access
cd /home/graham/workspace/experiments/chat # UI
```
Leave a review
No reviews yet. Be the first to review this skill!
# Download SKILL.md from killerskills.ai/api/skills/granger-ecosystem-context/raw