Chat With Your Bot From the Terminal

The three connection values

Quick terminal setup

1. 1

   Store the connection values

   Paste the Base URL and Model exactly as shown in Bash or Zsh. The hidden read keeps the API key out of your shell history, and the curl examples pass it through a temporary input stream instead of a process argument.

   Shell

   BOT_BASE_URL='PASTE_BASE_URL'
   BOT_MODEL='PASTE_MODEL'
   printf 'Endpoint API key: '
   read -s BOT_API_KEY
   printf '\n'

2. 2

   Check the connection

   Call the models route first. A successful models response means the URL, key, endpoint state, and running bot are all reachable. HTTP errors remain visible and make curl exit with a failure status.

   curl

   curl --fail-with-body -sS "$BOT_BASE_URL/models" \
     --header @<(printf 'Authorization: Bearer %s\n' "$BOT_API_KEY")

3. 3

   Send one message with curl

   The Base URL already contains /v1, so append only /chat/completions. This request returns the normal OpenAI-compatible JSON response.

   curl

   curl --fail-with-body -sS "$BOT_BASE_URL/chat/completions" \
     --header @<(printf 'Authorization: Bearer %s\n' "$BOT_API_KEY") \
     -H "Content-Type: application/json" \
     -d "{
       \"model\": \"$BOT_MODEL\",
       \"messages\": [
         {\"role\": \"user\", \"content\": \"Hello from my terminal\"}
       ]
     }"

   Print only the replyIf jq is installed, this Bash/Zsh variant preserves curl errors while printing only the reply:

   Bash / Zsh + jq

   set -o pipefail
   curl --fail-with-body -sS "$BOT_BASE_URL/chat/completions" \
     --header @<(printf 'Authorization: Bearer %s\n' "$BOT_API_KEY") \
     -H "Content-Type: application/json" \
     -d "{
       \"model\": \"$BOT_MODEL\",
       \"messages\": [
         {\"role\": \"user\", \"content\": \"Hello from my terminal\"}
       ]
     }" | jq -r '.choices[0].message.content'

4. 4

   Start an interactive terminal chat

   A single curl request does not remember earlier turns. This dependency-free Python script keeps the messages array in memory and resends it on every turn. Paste it into any terminal with Python 3.

   Python 3

   python3 -c "$(cat <<'PY'
   import json
   from getpass import getpass
   from urllib.error import HTTPError, URLError
   from urllib.request import Request, urlopen
   
   base_url = input("Base URL: ").strip().rstrip("/")
   model = input("Model: ").strip()
   api_key = getpass("API key: ")
   messages = []
   
   print("Chat ready. Use /new to clear history or /exit to quit.")
   
   while True:
       text = input("you> ").strip()
       if not text:
           continue
       if text in {"/exit", "/quit"}:
           break
       if text == "/new":
           messages.clear()
           print("History cleared.")
           continue
   
       messages.append({"role": "user", "content": text})
       payload = json.dumps({
           "model": model,
           "messages": messages,
           "stream": False,
       }).encode()
       request = Request(
           f"{base_url}/chat/completions",
           data=payload,
           headers={
               "Authorization": f"Bearer {api_key}",
               "Content-Type": "application/json",
           },
           method="POST",
       )
   
       try:
           with urlopen(request, timeout=300) as response:
               result = json.load(response)
           answer = result["choices"][0]["message"]["content"]
       except HTTPError as error:
           print(f"HTTP {error.code}: {error.read().decode(errors='replace')}")
           messages.pop()
           continue
       except (URLError, KeyError, IndexError, json.JSONDecodeError) as error:
           print(f"Request failed: {error}")
           messages.pop()
           continue
   
       print(f"bot> {answer}")
       messages.append({"role": "assistant", "content": answer})
   PY
   )"

unset BOT_API_KEY BOT_BASE_URL BOT_MODEL

Supported routes

The instance-card endpoint intentionally exposes only the two routes needed by compatible chat clients:

Troubleshooting

Related guides

• Connect the endpoint to TypingMind
• Use the browser Terminal and File Manager
• OpenClaw CLI command reference