Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.
Python agentaianthropicclaude-codecompressioncontext-engineeringcontext-windowcursorfastapilangchainllmmcpopenaiprompt-engineeringproxypythonragtoken-optimizationtokenstypescript
"""Actual OpenAI SDK HTTP requests: typed cache opt-out must preserve caller spaces."""
import hashlib
import json
import os
import tempfile
import threading
from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer
from pathlib import Path
def main():
with tempfile.TemporaryDirectory(prefix="headroom-cache-config-") as scratch:
for key in ["HEADROOM_WORKSPACE_DIR", "HEADROOM_CONFIG_DIR", "XDG_CACHE_HOME"]:
os.environ[key] = str(Path(scratch) / key.lower())
# This isolated proof needs no persistent, process-global CCR database.
# Keep the store in memory so Windows can remove the scratch directory.
os.environ.update(
HEADROOM_OFFLINE="1",
HEADROOM_BEACON="off",
DO_NOT_TRACK="1",
HEADROOM_CCR_BACKEND="memory",
)
from openai import OpenAI
from headroom import HeadroomClient, OpenAIProvider
from headroom.config import HeadroomConfig, HeadroomMode
requests = []
violations = []
class Handler(BaseHTTPRequestHandler):
def log_message(self, *args):
pass
def do_POST(self):
raw = self.rfile.read(int(self.headers["Content-Length"]))
body = json.loads(raw)
wire_dir = os.environ.get("HEADROOM_PROOF_WIRE_DIR")
if wire_dir:
target = Path(wire_dir)
target.mkdir(parents=True, exist_ok=True)
(target / f"request-{len(requests)}.json").write_bytes(raw)
requests.append(body)
payload = json.dumps(
{
"id": "chatcmpl_fixture",
"object": "chat.completion",
"created": 0,
"model": body["model"],
"choices": [
{
"index": 0,
"message": {"role": "assistant", "content": "local fixture"},
"finish_reason": "stop",
}
],
"usage": {"prompt_tokens": 1, "completion_tokens": 2, "total_tokens": 3},
}
).encode()
self.send_response(200)
self.send_header("Content-Type", "application/json")
self.send_header("Content-Length", str(len(payload)))
self.end_headers()
self.wfile.write(payload)
server = ThreadingHTTPServer(("127.0.0.1", 0), Handler)
threading.Thread(target=server.serve_forever, daemon=True).start()
original = OpenAI(
api_key="local-fixture",
base_url=f"http://127.0.0.1:{server.server_port}",
max_retries=0,
)
try:
for enabled in [False, True]:
url = f"sqlite:///{scratch}/{enabled}.db"
config = HeadroomConfig(
store_url=url,
default_mode=HeadroomMode.OPTIMIZE,
discover_pipeline_extensions=False,
)
config.smart_crusher.enabled = False
config.cache_aligner.enabled = False
config.cache_optimizer.enabled = enabled
client = HeadroomClient(original, OpenAIProvider(), store_url=url, config=config)
try:
response = client.chat.completions.create(
model="gpt-4o",
messages=[
{"role": "system", "content": "Keep exact spaces.\n"},
{"role": "user", "content": "Hello."},
],
max_tokens=8,
)
wire_system = requests[-1]["messages"][0]["content"]
print(
json.dumps(
{
"configured_enabled": enabled,
"runtime_optimizer": client.get_stats()["config"][
"cache_optimizer"
],
"wire_system": wire_system,
"wire_system_sha256": hashlib.sha256(
wire_system.encode()
).hexdigest(),
"response": response.choices[0].message.content,
}
)
)
if not enabled and wire_system != "Keep exact spaces.\n":
violations.append("Disabled cache optimizer modified the system message")
finally:
client.close()
finally:
original.close()
server.shutdown()
server.server_close()
assert not violations, violations
if __name__ == "__main__":
main()