optillm
Optimizing inference proxy for LLMs
Runssource: GitHubPythonApache-2.0commit eaf171aa6da5
Python, Apache-2.0 licensed. The project labels itself: agent, agentic ai, agentic framework, agentic workflow, agents, api gateway, chain of thought and genai.
optillm runs. An Argusic agent installed it in 18.5 minutes and hit 2 errors and fixed every one without help on a clean machine with no GPU, and the whole session was recorded.
OptiLLM package installed in a venv, plugins load correctly, CLI entry point works, test suite runs 153/163 passing (10 skipped due to MLX/environment conditions), and the local inference server starts, accepts HTTP requests, and generates real LLM responses with the codelion/dhara-250m model.
What the agent ended up with on a clean machine, in its own words. How this is measured
Measured by Argusic on a fresh machine every time. Every number links to its evidence. Argusic Score 60 of 100 (the mean of this project's run scores).
At a glance
Subject data from GitHub, linked at the top of this page, refreshed . Test data by Argusic (CC BY 4.0); every number links to a run page with the full log, the recording, and their sha256 hashes.
What was measured
From the run that decided this verdict. Every line is a recorded value, not an estimate.
What went wrong, in the agent's own words
The agent's own words, worst first, never rewritten into advice.
- mcp v2.2.0 missing mcp.client.websocket module2.5 minutes
- test_batching.py, test_reasoning_tokens.py can't find test_utils0.5 minutes
- Install time
- 19 minutes
- Cold machine to finish
- 22 minutes
- Errors hit and fixed
- 2 hit, 2 fixed with no human help
- How the result was proved
- curl -s -X POST http://localhost:8005/v1/chat/completions -H 'Content-Type: application/json' -H 'Authorization: Bearer optillm' -d '{"model":"codelion/dhara-250m","messages":[{"role":"user","content":"Hi"}],"max_tokens":10}' returned 200 with generated text; server health endpoint returned {"status":"ok"}
- Model tokens used
- 109,796
- Exact commit tested
- eaf171aa6da5
- Container image
- sha256:cdd920bce783
What was not measured
The other half, so you can see exactly how far this evidence goes. A gap is printed as a gap and never filled in.
- Peak GPU memory
- this ran on a CPU machine, so there was no GPU to sample
- Behaviour over a long run
- each attempt uses the smallest documented settings and a fixed time budget, so this measures whether it runs, not how it holds up under load
The end of the recording
The last lines of the terminal session that decided this verdict, exactly as they were recorded.
- `test_utils` not found from root dir in test_batching.py / test_reasoning_tokens.py → set `PYTHON… **Tests excluded** (need real API keys or a GPU): `test_api_compatibility.py`, `test_conversation_l… tokens used 109,796 Here's the summary of everything I did and found. **Install**: Created a venv, installed the package with `pip install -e .`. One dependency issue: t… **Tests run**: 153 passed, 10 skipped, 0 failed across all test suites that don't need external API… **Server verification (local inference)**: Started the server on port 8005 with `codelion/dhara-250… **Errors encountered and fixes**: - `mcp.client.websocket` not found in mcp v2.2.0 → installed mcp v1.30.0 - `test_utils` not found from root dir in test_batching.py / test_reasoning_tokens.py → set `PYTHON… **Tests excluded** (need real API keys or a GPU): `test_api_compatibility.py`, `test_conversation_l…
Replay the whole session, every command from a clean machine to this point.
How it was tested
One agent, one container, no human help. It installs the project the way its own documentation says to, runs it, and fixes what breaks. Everything below is recorded as it happened: the terminal session, the log and the exact commit. The full method.
Strengths and limits
Measured facts, not opinions. How this is written.
What went well
- Reached a running state on a clean machine, with the session recorded.
- Recovered from all 2 errors without a human stepping in, which says the failures are documented well enough to solve.
- Ran without a GPU, so it does not need one to start.
- Licensed Apache-2.0, as reported by its host.
- Measured 2 times, so the result is not a one-off.
What did not, or is not known
- Took 18.5 minutes to install, slower than the median of the 60 comparable projects Argusic has measured.
Of the 61 agent projects Argusic has installed and timed, optillm was the 49th fastest to reach a running state, and 42 of 61 reached one at all.
What it is a reasonable choice for
- Trying it on a laptop or a small server: it reached a running state without a GPU.
Also tested, in the same area
Every one of these was installed and run by Argusic on a clean machine. Nothing appears here that was not tested.
Run history
| tested by | attempt | outcome | score | cost USD | when |
|---|---|---|---|---|---|
| Argusic Runner | 2/3 | Runs | 100.00 | 0.11 | |
| Argusic Runner | 1/3 | Did not run | 20.00 | 0.20 |
Topics (from GitHub)
agentagentic-aiagentic-frameworkagentic-workflowagentsapi-gatewaychain-of-thoughtgenailarge-language-modelsllmllm-inferencellmapimixture-of-expertsmoamonte-carlo-tree-searchopenaiopenai-apioptimizationprompt-engineeringproxy-server
Embed the badge
Markdown for the project README. It links back here; terms on the terms page.
[](https://argusic.com/subject/optillm)Questions
- Does optillm run?
- Yes. optillm runs. Argusic installed and launched it on a clean machine in 19 minutes, hitting 2 errors on the way, and recorded the session.
- How did Argusic test optillm?
- On a fresh, disposable machine, with every command recorded and the repository pinned at commit eaf171aa6da5. 2 attempts are recorded, and the full method is on the methodology page.
- What did it cost to test optillm?
- The run that produced this verdict cost $0.11: real compute and model cost, not a list price per million tokens. It is recorded on that run's page.
- How long does optillm take to install?
- 18.5 minutes in the run that decided this verdict, from a clean machine to a running state. That is one measured attempt, not an average.
- Does optillm need a GPU?
- Not to start. Argusic reached a running state on a machine with no GPU.
- What goes wrong when installing optillm?
- 2 things broke in the recorded run, and 2 were fixed without human help. Each one, and the time it cost, is listed on this page.
- How does optillm compare with the alternatives?
- Of the 61 agent projects Argusic has installed and timed, optillm was the 49th fastest to reach a running state, and 42 of 61 reached one at all.
- Where is the evidence for optillm?
- All 2 recorded runs are on this page, each linking to its full log and terminal recording, stored with a sha256 fingerprint so it cannot be quietly altered.