https://github.com/maximerivest/lm15

Universal LM core with pluggable provider adapters for OpenAI, Anthropic, and Gemini.
https://github.com/maximerivest/lm15

adapters anthropic gemini llm openai python

Last synced: about 2 months ago
JSON representation

Universal LM core with pluggable provider adapters for OpenAI, Anthropic, and Gemini.

Host: GitHub
URL: https://github.com/maximerivest/lm15
Owner: MaximeRivest
License: mit
Created: 2026-04-08T17:14:38.000Z (about 2 months ago)
Default Branch: main
Last Pushed: 2026-04-10T16:52:20.000Z (about 2 months ago)
Last Synced: 2026-04-11T08:02:15.943Z (about 2 months ago)
Topics: adapters, anthropic, gemini, llm, openai, python
Language: Python
Homepage: https://pypi.org/project/lm15/
Size: 1.06 MB
Stars: 8
Watchers: 0
Forks: 0
Open Issues: 4
Metadata Files:
- Readme: README.md
- Changelog: CHANGELOG.md
- License: LICENSE

Awesome Lists containing this project

README

          # lm15

[![PyPI version](https://img.shields.io/pypi/v/lm15.svg)](https://pypi.org/project/lm15/)

[![Python 3.10+](https://img.shields.io/pypi/pyversions/lm15.svg)](https://pypi.org/project/lm15/)

[![MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](LICENSE)

One interface for OpenAI, Anthropic, and Gemini. Zero dependencies.

| | lm15 | google-genai | litellm |

|---|---:|---:|---:|

| **install** | 72ms | 137ms | 184ms |

| **import** | 95ms | 2,656ms | 4,534ms |

| **total (install → response)** | **1,090ms** | **3,992ms** | **5,840ms** |

| dependencies | 0 | 25 | 55 |

| disk footprint | 408K | 41M | 155M |

_{Median of 10 cold-start runs. Fresh venv, single completion against `gemini-3.1-flash-lite-preview`. [Benchmark source.](benchmarks/cold_start.sh)}

```python

import lm15

resp = lm15.complete("claude-sonnet-4-5", "Hello.")

print(resp.text)

```

Switch models by changing the string. Same types, same streaming, same tool calling. That's it.

> Yes, [we know](https://xkcd.com/927/).

## Install

```bash

pip install lm15

```

Set at least one provider key:

```bash

export OPENAI_API_KEY=sk-...

export ANTHROPIC_API_KEY=sk-ant-...

export GEMINI_API_KEY=...         # or GOOGLE_API_KEY

```

Discover what is available:

```python

import lm15

print(lm15.providers_info())

for m in lm15.models(provider="openai")[:5]:

    print(m.id)

```

## Usage

### Streaming

```python

for text in lm15.stream("gpt-4.1-mini", "Write a haiku.").text:

    print(text, end="")

```

Full event access:

```python

for event in lm15.stream("gpt-4.1-mini", "Write a haiku."):

    match event.type:

        case "text":     print(event.text, end="")

        case "thinking": print(f"💭 {event.text}", end="")

        case "finished": print(f"\n📊 {event.response.usage}")

```

### Tools (auto-execute)

Pass Python functions — schema is inferred, execution is automatic:

```python

def get_weather(city: str) -> str:

    """Get weather by city."""

    return f"22°C in {city}"

resp = lm15.complete("gpt-4.1-mini", "Weather in Montreal?", tools=[get_weather])

print(resp.text)  # "It's 22°C in Montreal."

```

### Tools (manual)

```python

from lm15 import Tool

weather = Tool(name="get_weather", description="Get weather", parameters={...})

gpt = lm15.model("gpt-4.1-mini")

resp = gpt("Weather in Montreal?", tools=[weather])

results = {tc.id: "22°C, sunny" for tc in resp.tool_calls}

resp = gpt.submit_tools(results)

print(resp.text)

```

### Images, audio, video, documents

```python

from lm15 import Part

# Image from URL

resp = lm15.complete("gemini-2.5-flash", ["Describe this.", Part.image(url="https://example.com/cat.jpg")])

# Image generation → vision (cross-model)

resp = lm15.complete("gpt-4.1-mini", "Draw a cat.", output="image")

resp2 = lm15.complete("claude-sonnet-4-5", ["What's this?", resp.image])

# Document

resp = lm15.complete("claude-sonnet-4-5", ["Summarize.", Part.document(url="https://example.com/paper.pdf")])

# Upload via provider file API

doc = lm15.upload("claude-sonnet-4-5", "contract.pdf")

resp = lm15.complete("claude-sonnet-4-5", ["Find liability clauses.", doc])

```

### Reasoning

```python

resp = lm15.complete("claude-sonnet-4-5", "Prove √2 is irrational.", reasoning=True)

print(resp.thinking)  # chain of thought

print(resp.text)      # final answer

```

### Conversation

```python

gpt = lm15.model("gpt-4.1-mini", system="You remember everything.")

gpt("My name is Max.")

gpt("I like chess.")

resp = gpt("What do you know about me?")

print(resp.text)  # knows both

```

### Prompt caching

Reduces cost and latency for repeated prefixes — system prompts, long documents, agent loops:

```python

agent = lm15.model("claude-sonnet-4-5",

    system="",

    tools=[read_file, write_file],

    prompt_caching=True,

)

resp = agent("Add tests for auth.")

while resp.finish_reason == "tool_call":

    results = execute(resp.tool_calls)

    resp = agent.submit_tools(results)

    print(f"Cache hit: {resp.usage.cache_read_tokens} tokens")

```

### Prefill

```python

resp = lm15.complete("claude-sonnet-4-5", "Output JSON for a person.", prefill="{")

```

### Reusable model with config

```python

gpt = lm15.model("gpt-4.1-mini", system="You are terse.", retries=3, cache=True, temperature=0)

resp = gpt("Hello.")

# Override per call

resp = gpt("Be creative.", temperature=1.5)

# Derive new models

claude = gpt.with_model("claude-sonnet-4-5")

```

### Config from dicts

```python

config = {"model": "gpt-4.1-mini", "system": "You are terse.", "temperature": 0}

resp = lm15.complete(prompt="Summarize DNA.", **config)

```

### Built-in tools

```python

resp = lm15.complete("gpt-4.1-mini", "Latest AI news", tools=["web_search"])

for c in resp.citations:

    print(c.title, c.url)

```

## Provider support

| Capability | OpenAI | Anthropic | Gemini |

|---|:---:|:---:|:---:|

| complete | ✅ | ✅ | ✅ |

| stream | ✅ | ✅ | ✅ |

| embeddings | ✅ | — | ✅ |

| files | ✅ | ✅ | ✅ |

| batches | ✅ | ✅ | ✅ |

| images | ✅ | — | ✅ |

| audio | ✅ | — | ✅ |

| prompt caching | auto | ✅ | ✅ |

## Architecture

```

lm15.complete / lm15.model       ← v2 surface (sugar)

        │

        ▼

LMRequest ──▶ UniversalLM ──▶ MiddlewarePipeline ──▶ ProviderAdapter ──▶ Transport

                  │                                        │

                  │ resolve_provider(model)                 │ build_request / parse_response

                  ▼                                        ▼

            capabilities.py                         providers/{openai,anthropic,gemini}.py

```

The v2 surface (`lm15.complete`, `lm15.model`, `Model`, `Stream`) is a thin layer that constructs `LMRequest` objects and calls `UniversalLM`. The universal provider contract is unchanged — third parties can build their own surface on top of the same internals.

## Why this exists

- **Stdlib only.** No `requests`, no `httpx`, no `aiohttp`. Transport is `urllib` or optional `pycurl`.

- **Frozen dataclasses all the way down.** `LMRequest` in, `LMResponse` out. No mutable builder chains.

- **Nothing is hidden.** Every internal type is importable. Provider escape hatches are always there.

- **Plugin discovery via entry points.** Third-party providers install and register without touching lm15 core.

## Docs

| Topic | Path |

|---|---|

| **API v2 spec** | [`docs/API_SPEC_V2.md`](docs/API_SPEC_V2.md) |

| Getting started | [`docs/GETTING_STARTED.md`](docs/GETTING_STARTED.md) |

| Core concepts | [`docs/CONCEPTS.md`](docs/CONCEPTS.md) |

| Architecture | [`docs/ARCHITECTURE.md`](docs/ARCHITECTURE.md) |

| Provider contract | [`docs/CONTRACT.md`](docs/CONTRACT.md) |

| Error handling | [`docs/ERRORS.md`](docs/ERRORS.md) |

| Streaming | [`docs/STREAMING.md`](docs/STREAMING.md) |

| Writing an adapter | [`docs/ADAPTER_GUIDE.md`](docs/ADAPTER_GUIDE.md) |

| Adding a provider | [`docs/ADD_PROVIDER_GUIDE.md`](docs/ADD_PROVIDER_GUIDE.md) |

| Completeness testing | [`docs/COMPLETENESS.md`](docs/COMPLETENESS.md) |

| Production checklist | [`docs/PRODUCTION_CHECKLIST.md`](docs/PRODUCTION_CHECKLIST.md) |

**Cookbooks v2:** [`docs/COOKBOOKS_V2/`](docs/COOKBOOKS_V2/) — practical examples + references:

1. [Hello World](docs/COOKBOOKS_V2/01-hello-world.md)

2. [Streaming](docs/COOKBOOKS_V2/02-streaming.md)

3. [Tools (auto-execute)](docs/COOKBOOKS_V2/03-tools-auto.md)

4. [Tools (manual loop)](docs/COOKBOOKS_V2/04-tools-manual.md)

5. [Multimodal](docs/COOKBOOKS_V2/05-multimodal.md)

6. [Reasoning](docs/COOKBOOKS_V2/06-reasoning.md)

7. [Conversation](docs/COOKBOOKS_V2/07-conversation.md)

8. [Prompt caching](docs/COOKBOOKS_V2/08-prompt-caching.md)

9. [Model config](docs/COOKBOOKS_V2/09-model-config.md)

10. [Building an agent](docs/COOKBOOKS_V2/10-agent.md)

11. [complete()/stream() reference](docs/COOKBOOKS_V2/11-complete-reference.md)

12. [Model discovery and provider status](docs/COOKBOOKS_V2/12-model-discovery.md)

**Cookbooks v1 (low-level):** [`docs/COOKBOOKS/`](docs/COOKBOOKS/) — 8 examples using the internal `LMRequest`/`UniversalLM` API directly.

## License

MIT

ecosyste.ms

Data

Tools

Indexes

Applications

Experiments

Awesome

https://github.com/maximerivest/lm15

Awesome Lists containing this project

README