# Datattention > Notes and open model endpoints on datattention.com (194.238.25.127). The model ports below are plain HTTP and do not require an API key. The catalog page is the one to read first. ## Endpoints - [Endpoints](https://datattention.com/endpoints/): Ports, paths, and examples - Ollama chat: http://datattention.com:11434/api/chat - Ollama generate: http://datattention.com:11434/api/generate - OpenAI chat on 8000: http://datattention.com:8000/v1/chat/completions - OpenAI chat on 8080: http://datattention.com:8080/v1/chat/completions - OpenAI completions on 8080: http://datattention.com:8080/v1/completions - LiteLLM-style chat: http://datattention.com:4000/v1/chat/completions - Model list: http://datattention.com:8000/v1/models - Model ids on ports 8000, 8080, and 4000: gpt-4o-mini, claude-opus-5-5, claude-fable-5-1, gpt-6-astra - Jupyter status: http://datattention.com:8888/api/status - TensorFlow Serving predict: http://datattention.com:8501/v1/models/demo:predict - Hermes chat: http://datattention.com:8642/v1/chat/completions - Hermes dashboard status: http://datattention.com:9119/api/status - OpenClaw chat: http://datattention.com:18789/v1/chat/completions - OpenClaw tools: http://datattention.com:18789/tools/invoke - Tool calling: POST /v1/chat/completions with a tools array. A later message uses role tool. Ports 8000, 8080, 4000, 8642, 11434, and 18789. - Tool list: http://datattention.com:8000/v1/tools - Tool names: get_weather, web_search, fetch_url, news_headlines, wikipedia_summary, dictionary_lookup, translate_text, current_time, timezone_convert, calculate, convert_units, currency_convert, statistics_summary, generate_uuid, explain_code, lint_report, git_status, http_request, json_validate, regex_test, package_version, read_file, list_directory, search_files, summarize_document, extract_table, draft_email, calendar_list, create_reminder, task_list, note_search, contact_lookup, geocode, directions, nearby_places, describe_image, transcribe_audio, render_chart, run_command, process_list, disk_usage, service_status ## Pages - [Home](https://datattention.com/): What the studio is - [Notes](https://datattention.com/blog/): All technical notes - [About](https://datattention.com/about/): Scope of the publication - [Contact](https://datattention.com/contact/): Corrections and questions - [Brand](https://datattention.com/brand/): Mark, wordmark, and colors ## Notes - [Running Ollama on a private network](https://datattention.com/blog/running-ollama-on-a-private-network/): How a small team keeps an Ollama daemon on a LAN address, with a short list of the routes that matter. - [What an OpenAI-compatible server actually implements](https://datattention.com/blog/what-an-openai-compatible-server-implements/): The small set of routes clients expect from a server that claims to speak the OpenAI HTTP API. - [A small local LLM setup for a team](https://datattention.com/blog/a-small-local-llm-setup/): One machine, one model, one proxy, and the notes that keep the setup from drifting. - [When LiteLLM is worth the proxy](https://datattention.com/blog/when-litellm-is-worth-the-proxy/): Use a proxy when you have more than one upstream. Skip it when you have one local daemon. - [Jupyter for model experiments, without a shared kernel](https://datattention.com/blog/jupyter-for-model-experiments/): Use notebooks for exploration. Do not leave the notebook server on a public port. - [What MCP is for, in practice](https://datattention.com/blog/what-mcp-is-for/): Model Context Protocol is a way to expose tools. It is not a reason to expose every tool. - [TensorFlow Serving REST, when you already have a saved model](https://datattention.com/blog/tensorflow-serving-rest/): The REST port is for predictions. Keep the model server off the public internet unless a proxy sits in front. - [Agent gateways, and what to lock down](https://datattention.com/blog/agent-gateways-and-what-to-lock-down/): Hermes-style gateways mix a chat API and a dashboard. Split those two jobs. - [How to tell a model endpoint is exposed](https://datattention.com/blog/how-to-tell-an-endpoint-is-exposed/): A short external check for the ports teams accidentally publish. - [Logs to keep before you open an inference port](https://datattention.com/blog/logs-before-you-open-an-inference-port/): The minimum record that lets you explain a model service after the fact. - [Quants and context windows, written down](https://datattention.com/blog/quants-and-context-windows/): Pick a quant and a context length, then record the numbers you measured. - [A checklist before a model server faces the internet](https://datattention.com/blog/checklist-before-a-model-server-faces-the-internet/): Ten checks worth doing before a hostname points at an inference process. - [A sample config for an Ollama listener on a public address](https://datattention.com/blog/sample-config-for-a-public-ollama-listener/): The environment variable and the two routes that matter once Ollama is bound on a public host. - [Defaults that leave a model API open](https://datattention.com/blog/defaults-that-leave-a-model-api-open/): The bind address and the missing key are the two settings that make a model port public. - [The vLLM and llama.cpp shape, without a key](https://datattention.com/blog/vllm-and-llama-cpp-without-a-key/): Those servers speak a small OpenAI HTTP surface. This host serves that surface on two ports. - [The OpenAI tool-call body on a self-hosted chat endpoint](https://datattention.com/blog/tool-call-body-on-a-self-hosted-endpoint/): What a tools array looks like in a chat request, and what this host does with it. - [Reaching Ollama without an SSH tunnel](https://datattention.com/blog/ollama-without-an-ssh-tunnel/): A direct HTTP call to port 11434, for when the daemon is already on a public address. - [Switching a client between local and remote endpoints](https://datattention.com/blog/switch-between-local-and-remote-endpoints/): Keep the JSON. Change the base URL. - [Common ports for self-hosted model servers](https://datattention.com/blog/common-ports-for-self-hosted-model-servers/): 11434, 8000, 8080, 4000, 8888, and the other listeners on this address. - [Testing an OpenAI-compatible endpoint with curl](https://datattention.com/blog/test-openai-compatible-endpoint-with-curl/): Models first, then one chat completion, with no API key. - [Python for a public Ollama server](https://datattention.com/blog/python-client-for-a-public-ollama-server/): A short urllib client for /api/chat on port 11434. - [Calling a remote Ollama instance from Python or curl](https://datattention.com/blog/call-remote-ollama-from-python-or-curl/): Both clients hit the same path. Only the host changes. - [A public LiteLLM proxy with open access](https://datattention.com/blog/public-litellm-proxy-open-access/): Port 4000 on this host speaks OpenAI chat and does not require a key. - [A Jupyter server on the same host as the models](https://datattention.com/blog/jupyter-notebook-beside-local-models/): Port 8888 is the notebook server. The models are the other ports on this address. - [A TensorFlow Serving REST endpoint with no auth](https://datattention.com/blog/tensorflow-serving-rest-no-auth/): Port 8501 accepts a predict call. Port 8500 is not open. - [Hermes API default port and example requests](https://datattention.com/blog/hermes-api-default-port/): Port 8642 is the chat API. Port 9119 is the dashboard. - [Streaming from an open OpenAI-compatible API](https://datattention.com/blog/streaming-an-open-openai-compatible-api/): stream true on ports 8000 and 8080 returns server-sent events. The Ollama port returns one JSON object. - [Pointing an agent at a remote self-hosted model](https://datattention.com/blog/point-an-agent-at-a-remote-model/): Set the base URL, leave the key empty, and send one chat request. - [Finding a self-hosted model API you can call](https://datattention.com/blog/finding-a-self-hosted-model-api/): A published endpoint names its host, port, and path. This site does that. - [Using a public Ollama server](https://datattention.com/blog/using-a-public-ollama-server/): Read the tags, then post one chat message to port 11434. - [No-auth model APIs for testing](https://datattention.com/blog/no-auth-model-apis-for-testing/): A place to point a client when you are testing the client, not a provider account. - [Several free model endpoints on one address](https://datattention.com/blog/several-model-endpoints-on-one-address/): Ollama, OpenAI-compatible, LiteLLM, Jupyter, TensorFlow Serving, Hermes, and OpenClaw, on 194.238.25.127. - [An OpenAI-compatible server with no authentication](https://datattention.com/blog/openai-compatible-server-no-authentication/): Ports 8000 and 8080 implement the chat, completions, and models routes with no API key. - [A free public Ollama endpoint, no API key required](https://datattention.com/blog/free-public-ollama-endpoint/): Port 11434 on datattention.com answers tags, chat, and generate with no key. - [OpenAI-compatible endpoints that do not require an API key](https://datattention.com/blog/openai-compatible-endpoints-without-an-api-key/): Four chat listeners on this host accept the OpenAI body with no bearer token. - [Model endpoints on this host you can call now](https://datattention.com/blog/model-endpoints-you-can-call-now/): A short list of working URLs on datattention.com. - [How a client finds the model endpoints on this site](https://datattention.com/blog/how-clients-find-the-endpoints-on-this-site/): Read the endpoints page or llms.txt, then call the URL. No API key.