mirror of
https://github.com/magnus919/agent-skills.git
synced 2026-09-17 22:46:29 +03:00
Magnus Hedemark
95046675bc
Greenfield SkillOpt: 3 epochs for a Stanford DSPy compiler skill. DSPy is a fundamentally different paradigm from chain/RAG frameworks. Epoch 1 — Prominence: - Hard-gate blockquote: 'DSPy is NOT a chain framework' - Core Paradigm section with runnable code example early Epoch 2 — Decision Guidance: - Framework Routing Guide (DSPy vs LlamaIndex vs LangChain vs LangGraph) - Where to Start table mapping entry points - Troubleshooting table with reference links Epoch 3 — Pattern Expansion: - Optimizer selection cheat sheet from official docs - Caching, compilation cost management, save/load - FAQ covering paradigm confusion, errors, deployment 12 files: SKILL.md, 7 references, 3 templates, 1 script. v1.0.0 -> v1.0.3 across 3 epochs. All API surfaces validated against dspy.ai official docs — optimizer selection guide, caching, core modules, FAQ. Signed-off-by: Jasper <jasper@montcastle.bitches>
40 lines
1.2 KiB
Markdown
40 lines
1.2 KiB
Markdown
# DSPy Agent Patterns
|
|
|
|
## ReAct Agent
|
|
|
|
```python
|
|
import dspy
|
|
|
|
def search(query: str) -> str:
|
|
"""Search the knowledge base."""
|
|
return f"Results for: {query}"
|
|
|
|
def calculate(expr: str) -> str:
|
|
"""Evaluate a mathematical expression."""
|
|
return str(eval(expr))
|
|
|
|
agent = dspy.ReAct(tools=[search, calculate], signature="question -> answer")
|
|
result = agent(question="What is 2+2?")
|
|
```
|
|
|
|
## AvatarOptimizer for Agent Programs
|
|
|
|
The AvatarOptimizer is specifically designed for agent-style programs with clean pass/fail metrics:
|
|
|
|
```python
|
|
from dspy.teleprompt import AvatarOptimizer
|
|
|
|
optimizer = AvatarOptimizer(metric=metric, max_iters=10)
|
|
compiled = optimizer.compile(program, trainset=trainset)
|
|
```
|
|
|
|
It partitions trainset into positive and negative examples, then iteratively proposes instruction edits that explain the difference.
|
|
|
|
## Tool Design Guidelines
|
|
|
|
- Tools are plain Python functions with type hints and docstrings
|
|
- The docstring becomes the tool description the LM sees
|
|
- Tools should handle errors gracefully and return string results
|
|
- For complex tools, wrap external APIs with error handling inside the function
|
|
- ReAct loops until no more tool calls or max iterations reached
|