mirror of
https://github.com/magnus919/agent-skills.git
synced 2026-09-15 21:46:29 +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>
2.3 KiB
2.3 KiB
DSPy Core Modules
DSPy modules are the building blocks of prompt programs. Unlike LangChain chains, modules are compiled — the optimizer tunes their prompts automatically.
dspy.Predict
The simplest module. Makes a direct prediction without intermediate reasoning.
import dspy
class Sentiment(dspy.Signature):
"""Classify sentiment."""
text: str = dspy.InputField()
sentiment: str = dspy.OutputField(desc="positive, negative, or neutral")
classifier = dspy.Predict(Sentiment)
result = classifier(text="DSPy is amazing!")
print(result.sentiment) # "positive"
dspy.ChainOfThought
Adds step-by-step reasoning before answering. Almost always beats Predict on complex tasks.
qa = dspy.ChainOfThought("question, context -> answer")
result = qa(question="What is the capital?", context="France is a country in Europe.")
print(f"Reasoning: {result.reasoning}")
print(f"Answer: {result.answer}")
dspy.ReAct — Tool-Use Agent
Combines reasoning with tool calls in a loop. Takes tools as callable functions.
def search(query: str) -> str:
"""Search the web."""
return f"Results for: {query}"
# ReAct accepts a list of tools
agent = dspy.ReAct(tools=[search], signature="question -> answer")
result = agent(question="What is the latest DSPy version?")
print(result.answer)
Custom dspy.Module
For multi-step programs with Python control flow:
class RAG(dspy.Module):
def __init__(self):
self.retrieve = dspy.Retrieve(k=5)
self.generate = dspy.ChainOfThought("context, question -> answer")
def forward(self, question):
context = self.retrieve(question).passages
return self.generate(question=question, context=context)
rag = RAG()
result = rag("What is DSPy?")
dspy.Assert and dspy.Suggest
Validation within programs:
class VerifiedQA(dspy.Module):
def __init__(self):
self.qa = dspy.ChainOfThought("question -> answer")
def forward(self, question):
pred = self.qa(question=question)
dspy.Suggest(len(pred.answer) > 10, "Answer should be detailed.")
return pred
# Activate assertions
from dspy.primitives.assertions import assert_transform_module, backtrack_handler
verified = assert_transform_module(VerifiedQA(), backtrack_handler)