Self-Host Laguna S 2.1 as an OpenAI-Compatible Coding API

pip install vllm
# Two-GPU serving with FP8 quantization and a working context cap.
# Raise --max-model-len only when a task actually needs it; KV cache
# for 1M tokens is its own capacity project.
vllm serve poolside/Laguna-S-2.1 \
--quantization fp8 \
--tensor-parallel-size 2 \
--max-model-len 131072 \
--enable-auto-tool-choice \
--tool-call-parser hermes \
--port 8000

curl -s http://localhost:8000/v1/models | python3 -m json.tool

from openai import OpenAI
client = OpenAI(base_url="http://localhost:8000/v1", api_key="local")
response = client.chat.completions.create(
model="poolside/Laguna-S-2.1",
messages=[
{"role": "system", "content": "You are a precise coding assistant."},
{"role": "user", "content": "Write a Python function that parses "
"RFC 3339 timestamps without external dependencies."},
],
temperature=0.2,
max_tokens=1024,
)
print(response.choices[0].message.content)

import json
import subprocess
from pathlib import Path
REPO_ROOT = Path("/srv/repos/target-repo").resolve()
TOOLS = [
{
"type": "function",
"function": {
"name": "read_file",
"description": "Read a file from the repository.",
"parameters": {
"type": "object",
"properties": {"path": {"type": "string"}},
"required": ["path"],
},
},
},
{
"type": "function",
"function": {
"name": "run_command",
"description": "Run a read-only git or test command.",
"parameters": {
"type": "object",
"properties": {"command": {"type": "string"}},
"required": ["command"],
},
},
},
]
ALLOWED_PREFIXES = ("git diff", "git log", "git show", "pytest", "python -m pytest")
def read_file(path: str) -> str:
target = (REPO_ROOT / path).resolve()
if not target.is_relative_to(REPO_ROOT):
return "ERROR: path escapes the repository root"
return target.read_text()[:20000]
def run_command(command: str) -> str:
if not command.startswith(ALLOWED_PREFIXES):
return f"ERROR: command not in allowlist: {command}" 
result = subprocess.run(
command.split(), cwd=REPO_ROOT, capture_output=True, text=True, timeout=120
)
return (result.stdout + result.stderr)[:20000]

def run_agent(task: str, max_steps: int = 12) -> str:
messages = [
{"role": "system", "content": (
"You are a code reviewer. Use read_file and run_command to inspect "
"the repository before answering. Cite file paths and line numbers."
)},
{"role": "user", "content": task},
]
for _ in range(max_steps):
response = client.chat.completions.create(
model="poolside/Laguna-S-2.1",
messages=messages,
tools=TOOLS,
temperature=0.2,
)
msg = response.choices[0].message
if not msg.tool_calls:
return msg.content
messages.append(msg)
for call in msg.tool_calls:
args = json.loads(call.function.arguments)
output = (read_file(**args) if call.function.name == "read_file"
else run_command(**args))
messages.append({
"role": "tool", "tool_call_id": call.id, "content": output,
})
return "ERROR: step budget exhausted before the agent finished"
print(run_agent("Review the diff on the current branch: git diff main...HEAD. "
"Flag correctness issues only, ranked by severity."))

"""PR-diff reviewer on self-hosted poolside Laguna S 2.1 via vLLM.
Start the server first:
vllm serve poolside/Laguna-S-2.1 --quantization fp8 \
--tensor-parallel-size 2 --max-model-len 131072 \
--enable-auto-tool-choice --tool-call-parser hermes --port 8000
"""
import json
import subprocess
from pathlib import Path
from openai import OpenAI
REPO_ROOT = Path("/srv/repos/target-repo").resolve()
ALLOWED_PREFIXES = ("git diff", "git log", "git show", "pytest", "python -m pytest")
client = OpenAI(base_url="http://localhost:8000/v1", api_key="local")
TOOLS = [
{"type": "function", "function": {
"name": "read_file",
"description": "Read a file from the repository.",
"parameters": {"type": "object",
"properties": {"path": {"type": "string"}},
"required": ["path"]}}},
{"type": "function", "function": {
"name": "run_command",
"description": "Run a read-only git or test command.",
"parameters": {"type": "object",
"properties": {"command": {"type": "string"}},
"required": ["command"]}}},
]
def read_file(path: str) -> str:
target = (REPO_ROOT / path).resolve()
if not target.is_relative_to(REPO_ROOT):
return "ERROR: path escapes the repository root"
return target.read_text()[:20000]
def run_command(command: str) -> str:
if not command.startswith(ALLOWED_PREFIXES):
return f"ERROR: command not in allowlist: {command}"
result = subprocess.run(command.split(), cwd=REPO_ROOT,
capture_output=True, text=True, timeout=120)
return (result.stdout + result.stderr)[:20000]
def run_agent(task: str, max_steps: int = 12) -> str:
messages = [
{"role": "system", "content": (
"You are a code reviewer. Use read_file and run_command to inspect "
"the repository before answering. Cite file paths and line numbers.")},
{"role": "user", "content": task},
]
for _ in range(max_steps):
response = client.chat.completions.create(
model="poolside/Laguna-S-2.1", messages=messages,
tools=TOOLS, temperature=0.2)
msg = response.choices[0].message
if not msg.tool_calls:
return msg.content
messages.append(msg)
for call in msg.tool_calls:
args = json.loads(call.function.arguments)
output = (read_file(**args) if call.function.name == "read_file"
else run_command(**args))
messages.append({"role": "tool", "tool_call_id": call.id,
"content": output})
return "ERROR: step budget exhausted before the agent finished"
if __name__ == "__main__":
print(run_agent("Review the diff on the current branch: "
"git diff main...HEAD. Flag correctness issues only, "
"ranked by severity."))