outlines

Outlines：结构化文本生成

何时使用本技能

在以下场景中使用 Outlines：

• 在生成过程中保证 JSON/XML/代码结构有效
• 使用 Pydantic 模型实现类型安全的输出
• 支持本地模型（Transformers、llama.cpp、vLLM）
• 通过零开销的结构化生成最大化推理速度
• 根据 JSON schema 自动生成
• 在语法层面控制 token 采样

GitHub Stars：8,000+ | 来自：dottxt.ai（原 .txt）

安装

# Base installation
pip install outlines

# With specific backends
pip install outlines transformers  # Hugging Face models
pip install outlines llama-cpp-python  # llama.cpp
pip install outlines vllm  # vLLM for high-throughput

快速开始

基础示例：分类

import outlines
from typing import Literal

# Load model
model = outlines.models.transformers("microsoft/Phi-3-mini-4k-instruct")

# Generate with type constraint
prompt = "Sentiment of 'This product is amazing!': "
generator = outlines.generate.choice(model, ["positive", "negative", "neutral"])
sentiment = generator(prompt)

print(sentiment)  # "positive" (guaranteed one of these)

使用 Pydantic 模型

from pydantic import BaseModel
import outlines

class User(BaseModel):
    name: str
    age: int
    email: str

model = outlines.models.transformers("microsoft/Phi-3-mini-4k-instruct")

# Generate structured output
prompt = "Extract user: John Doe, 30 years old, [email protected]"
generator = outlines.generate.json(model, User)
user = generator(prompt)

print(user.name)   # "John Doe"
print(user.age)    # 30
print(user.email)  # "[email protected]"

核心概念

1. 约束式 token 采样

Outlines 使用有限状态机（FSM）在 logit 层面约束 token 生成。

工作原理：

1. 将 schema（JSON/Pydantic/正则表达式）转换为上下文无关文法（CFG）
2. 将 CFG 转换为有限状态机（FSM）
3. 在生成过程中的每一步过滤无效 token
4. 当仅存在一个有效 token 时快速推进

优势：

• 零开销：过滤在 token 层面进行
• 速度提升：沿确定性路径快速推进
• 保证有效性：无效输出不可能出现

import outlines

# Pydantic model -> JSON schema -> CFG -> FSM
class Person(BaseModel):
    name: str
    age: int

model = outlines.models.transformers("microsoft/Phi-3-mini-4k-instruct")

# Behind the scenes:
# 1. Person -> JSON schema
# 2. JSON schema -> CFG
# 3. CFG -> FSM
# 4. FSM filters tokens during generation

generator = outlines.generate.json(model, Person)
result = generator("Generate person: Alice, 25")

2. 结构化生成器

Outlines 为不同输出类型提供专用生成器。

选择生成器

# Multiple choice selection
generator = outlines.generate.choice(
    model,
    ["positive", "negative", "neutral"]
)

sentiment = generator("Review: This is great!")
# Result: One of the three choices

JSON 生成器

from pydantic import BaseModel

class Product(BaseModel):
    name: str
    price: float
    in_stock: bool

# Generate valid JSON matching schema
generator = outlines.generate.json(model, Product)
product = generator("Extract: iPhone 15, $999, available")

# Guaranteed valid Product instance
print(type(product))  # <class '__main__.Product'>

正则表达式生成器

# Generate text matching regex
generator = outlines.generate.regex(
    model,
    r"[0-9]{3}-[0-9]{3}-[0-9]{4}"  # Phone number pattern
)

phone = generator("Generate phone number:")
# Result: "555-123-4567" (guaranteed to match pattern)

整数/浮点数生成器

# Generate specific numeric types
int_generator = outlines.generate.integer(model)
age = int_generator("Person's age:")  # Guaranteed integer

float_generator = outlines.generate.float(model)
price = float_generator("Product price:")  # Guaranteed float

3. 模型后端

Outlines 支持多种本地和基于 API 的后端。

Transformers（Hugging Face）

import outlines

# Load from Hugging Face
model = outlines.models.transformers(
    "microsoft/Phi-3-mini-4k-instruct",
    device="cuda"  # Or "cpu"
)

# Use with any generator
generator = outlines.generate.json(model, YourModel)

llama.cpp

# Load GGUF model
model = outlines.models.llamacpp(
    "./models/llama-3.1-8b-instruct.Q4_K_M.gguf",
    n_gpu_layers=35
)

generator = outlines.generate.json(model, YourModel)

vLLM（高吞吐量）

# For production deployments
model = outlines.models.vllm(
    "meta-llama/Llama-3.1-8B-Instruct",
    tensor_parallel_size=2  # Multi-GPU
)

generator = outlines.generate.json(model, YourModel)

OpenAI（有限支持）

# Basic OpenAI support
model = outlines.models.openai(
    "gpt-4o-mini",
    api_key="your-api-key"
)

# Note: Some features limited with API models
generator = outlines.generate.json(model, YourModel)

4. Pydantic 集成

Outlines 提供一等的 Pydantic 支持，并可自动转换 schema。

基础模型

from pydantic import BaseModel, Field

class Article(BaseModel):
    title: str = Field(description="Article title")
    author: str = Field(description="Author name")
    word_count: int = Field(description="Number of words", gt=0)
    tags: list[str] = Field(description="List of tags")

model = outlines.models.transformers("microsoft/Phi-3-mini-4k-instruct")
generator = outlines.generate.json(model, Article)

article = generator("Generate article about AI")
print(article.title)
print(article.word_count)  # Guaranteed > 0

嵌套模型

class Address(BaseModel):
    street: str
    city: str
    country: str

class Person(BaseModel):
    name: str
    age: int
    address: Address  # Nested model

generator = outlines.generate.json(model, Person)
person = generator("Generate person in New York")

print(person.address.city)  # "New York"

枚举和字面量

from enum import Enum
from typing import Literal

class Status(str, Enum):
    PENDING = "pending"
    APPROVED = "approved"
    REJECTED = "rejected"

class Application(BaseModel):
    applicant: str
    status: Status  # Must be one of enum values
    priority: Literal["low", "medium", "high"]  # Must be one of literals

generator = outlines.generate.json(model, Application)
app = generator("Generate application")

print(app.status)  # Status.PENDING (or APPROVED/REJECTED)

常见模式

模式 1：数据提取

from pydantic import BaseModel
import outlines

class CompanyInfo(BaseModel):
    name: str
    founded_year: int
    industry: str
    employees: int

model = outlines.models.transformers("microsoft/Phi-3-mini-4k-instruct")
generator = outlines.generate.json(model, CompanyInfo)

text = """
Apple Inc. was founded in 1976 in the technology industry.
The company employs approximately 164,000 people worldwide.
"""

prompt = f"Extract company information:\n{text}\n\nCompany:"
company = generator(prompt)

print(f"Name: {company.name}")
print(f"Founded: {company.founded_year}")
print(f"Industry: {company.industry}")
print(f"Employees: {company.employees}")

模式 2：分类

from typing import Literal
import outlines

model = outlines.models.transformers("microsoft/Phi-3-mini-4k-instruct")

# Binary classification
generator = outlines.generate.choice(model, ["spam", "not_spam"])
result = generator("Email: Buy now! 50% off!")

# Multi-class classification
categories = ["technology", "business", "sports", "entertainment"]
category_gen = outlines.generate.choice(model, categories)
category = category_gen("Article: Apple announces new iPhone...")

# With confidence
class Classification(BaseModel):
    label: Literal["positive", "negative", "neutral"]
    confidence: float

classifier = outlines.generate.json(model, Classification)
result = classifier("Review: This product is okay, nothing special")

模式 3：结构化表单

class UserProfile(BaseModel):
    full_name: str
    age: int
    email: str
    phone: str
    country: str
    interests: list[str]

model = outlines.models.transformers("microsoft/Phi-3-mini-4k-instruct")
generator = outlines.generate.json(model, UserProfile)

prompt = """
Extract user profile from:
Name: Alice Johnson
Age: 28
Email: [email protected]
Phone: 555-0123
Country: USA
Interests: hiking, photography, cooking
"""

profile = generator(prompt)
print(profile.full_name)
print(profile.interests)  # ["hiking", "photography", "cooking"]

模式 4：多实体提取

class Entity(BaseModel):
    name: str
    type: Literal["PERSON", "ORGANIZATION", "LOCATION"]

class DocumentEntities(BaseModel):
    entities: list[Entity]

model = outlines.models.transformers("microsoft/Phi-3-mini-4k-instruct")
generator = outlines.generate.json(model, DocumentEntities)

text = "Tim Cook met with Satya Nadella at Microsoft headquarters in Redmond."
prompt = f"Extract entities from: {text}"

result = generator(prompt)
for entity in result.entities:
    print(f"{entity.name} ({entity.type})")

模式 5：代码生成

class PythonFunction(BaseModel):
    function_name: str
    parameters: list[str]
    docstring: str
    body: str

model = outlines.models.transformers("microsoft/Phi-3-mini-4k-instruct")
generator = outlines.generate.json(model, PythonFunction)

prompt = "Generate a Python function to calculate factorial"
func = generator(prompt)

print(f"def {func.function_name}({', '.join(func.parameters)}):")
print(f'    """{func.docstring}"""')
print(f"    {func.body}")

模式 6：批处理

def batch_extract(texts: list[str], schema: type[BaseModel]):
    """Extract structured data from multiple texts."""
    model = outlines.models.transformers("microsoft/Phi-3-mini-4k-instruct")
    generator = outlines.generate.json(model, schema)

    results = []
    for text in texts:
        result = generator(f"Extract from: {text}")
        results.append(result)

    return results

class Person(BaseModel):
    name: str
    age: int

texts = [
    "John is 30 years old",
    "Alice is 25 years old",
    "Bob is 40 years old"
]

people = batch_extract(texts, Person)
for person in people:
    print(f"{person.name}: {person.age}")

后端配置

Transformers

import outlines

# Basic usage
model = outlines.models.transformers("microsoft/Phi-3-mini-4k-instruct")

# GPU configuration
model = outlines.models.transformers(
    "microsoft/Phi-3-mini-4k-instruct",
    device="cuda",
    model_kwargs={"torch_dtype": "float16"}
)

# Popular models
model = outlines.models.transformers("meta-llama/Llama-3.1-8B-Instruct")
model = outlines.models.transformers("mistralai/Mistral-7B-Instruct-v0.3")
model = outlines.models.transformers("Qwen/Qwen2.5-7B-Instruct")

llama.cpp

# Load GGUF model
model = outlines.models.llamacpp(
    "./models/llama-3.1-8b.Q4_K_M.gguf",
    n_ctx=4096,         # Context window
    n_gpu_layers=35,    # GPU layers
    n_threads=8         # CPU threads
)

# Full GPU offload
model = outlines.models.llamacpp(
    "./models/model.gguf",
    n_gpu_layers=-1  # All layers on GPU
)

vLLM（生产环境）

# Single GPU
model = outlines.models.vllm("meta-llama/Llama-3.1-8B-Instruct")

# Multi-GPU
model = outlines.models.vllm(
    "meta-llama/Llama-3.1-70B-Instruct",
    tensor_parallel_size=4  # 4 GPUs
)

# With quantization
model = outlines.models.vllm(
    "meta-llama/Llama-3.1-8B-Instruct",
    quantization="awq"  # Or "gptq"
)

最佳实践

1. 使用具体类型

# ✅ Good: Specific types
class Product(BaseModel):
    name: str
    price: float  # Not str
    quantity: int  # Not str
    in_stock: bool  # Not str

# ❌ Bad: Everything as string
class Product(BaseModel):
    name: str
    price: str  # Should be float
    quantity: str  # Should be int

2. 添加约束

from pydantic import Field

# ✅ Good: With constraints
class User(BaseModel):
    name: str = Field(min_length=1, max_length=100)
    age: int = Field(ge=0, le=120)
    email: str = Field(pattern=r"^[\w\.-]+@[\w\.-]+\.\w+$")

# ❌ Bad: No constraints
class User(BaseModel):
    name: str
    age: int
    email: str

3. 为类别使用枚举

# ✅ Good: Enum for fixed set
class Priority(str, Enum):
    LOW = "low"
    MEDIUM = "medium"
    HIGH = "high"

class Task(BaseModel):
    title: str
    priority: Priority

# ❌ Bad: Free-form string
class Task(BaseModel):
    title: str
    priority: str  # Can be anything

4. 在提示词中提供上下文

# ✅ Good: Clear context
prompt = """
Extract product information from the following text.
Text: iPhone 15 Pro costs $999 and is currently in stock.
Product:
"""

# ❌ Bad: Minimal context
prompt = "iPhone 15 Pro costs $999 and is currently in stock."

5. 处理可选字段

from typing import Optional

# ✅ Good: Optional fields for incomplete data
class Article(BaseModel):
    title: str  # Required
    author: Optional[str] = None  # Optional
    date: Optional[str] = None  # Optional
    tags: list[str] = []  # Default empty list

# Can succeed even if author/date missing

与其他方案对比

功能	Outlines	Instructor	Guidance	LMQL
Pydantic 支持	✅ 原生	✅ 原生	❌ 无	❌ 无
JSON Schema	✅ 支持	✅ 支持	⚠️ 有限	✅ 支持
正则表达式约束	✅ 支持	❌ 不支持	✅ 支持	✅ 支持
本地模型	✅ 完整	⚠️ 有限	✅ 完整	✅ 完整
API 模型	⚠️ 有限	✅ 完整	✅ 完整	✅ 完整
零开销	✅ 支持	❌ 不支持	⚠️ 部分	✅ 支持
自动重试	❌ 不支持	✅ 支持	❌ 不支持	❌ 不支持
学习曲线	低	低	低	高

何时选择 Outlines：

• 使用本地模型（Transformers、llama.cpp、vLLM）
• 需要最大推理速度
• 希望获得 Pydantic 模型支持
• 要求零开销的结构化生成
• 控制 token 采样过程

何时选择其他方案：

• Instructor：需要支持自动重试的 API 模型
• Guidance：需要 token 修复和复杂工作流
• LMQL：偏好声明式查询语法

性能特征

速度：

• 零开销：结构化生成与无约束生成一样快
• 快速推进优化：跳过确定性 token
• 比生成后验证方法快 1.2-2 倍

内存：

• 每个 schema 编译一次 FSM（已缓存）
• 运行时开销极小
• 配合 vLLM 可实现高吞吐量且高效

准确性：

• 100% 有效输出（由 FSM 保证）
• 无需重试循环
• 确定性的 token 过滤

资源

• 文档：https://outlines-dev.github.io/outlines
• GitHub：https://github.com/outlines-dev/outlines（8k+ star）
• Discord：https://discord.gg/R9DSu34mGd
• 博客：https://blog.dottxt.co

另请参阅

• references/json_generation.md - 全面的 JSON 与 Pydantic 模式
• references/backends.md - 后端专属配置
• references/examples.md - 可直接用于生产的示例