langchain-ai/deepagents · error · ValueError

--model-params requires a JSON object value

Error message

--model-params requires a JSON object value

What it means

_extract_model_params_flag parses the value following the --model-params CLI flag. When there is nothing after the flag (only whitespace or end of input), it cannot build a JSON object and raises ValueError. The flag requires a JSON object payload such as {"temperature":0.2}.

Source

Thrown at libs/code/deepagents_code/app.py:1934

        Tuple of `(remaining_args, parsed_dict | None)`. Returns `None` for the
            dict when the flag is absent.

    Raises:
        ValueError: If the value is missing, has unclosed quotes,
            unbalanced braces, or is not valid JSON.
        TypeError: If the parsed JSON is not a dict.
    """
    flag = "--model-params"
    idx = raw_arg.find(flag)
    if idx == -1:
        return raw_arg, None

    before = raw_arg[:idx].rstrip()
    after = raw_arg[idx + len(flag) :].lstrip()

    if not after:
        msg = "--model-params requires a JSON object value"
        raise ValueError(msg)

    # Determine the JSON string boundaries.
    if after[0] in {"'", '"'}:
        quote = after[0]
        end = -1
        backslash_count = 0
        for i, ch in enumerate(after[1:], start=1):
            if ch == "\\":
                backslash_count += 1
                continue
            if ch == quote and backslash_count % 2 == 0:
                end = i
                break
            backslash_count = 0
        if end == -1:
            msg = f"Unclosed {quote} in --model-params value"
            raise ValueError(msg)
        # Parse the quoted token with shlex so escaped quotes are unescaped.

View on GitHub (pinned to a1af029e6e)

Solutions

1. Append a JSON object after the flag: --model-params '{"temperature":0.2}'.
2. Quote the JSON if your shell strips braces, and verify the variable feeding the flag is non-empty.
3. If the value comes from a script, echo it first to confirm it expands to a JSON object.

Example fix

# before
dcode --model-params

# after
dcode --model-params '{"temperature": 0.2}'

parts = shlex.split(cmd)
if "--model-params" in parts and parts.index("--model-params") == len(parts) - 1:
    raise SystemExit("--model-params needs a JSON object value")

try:
    handle_command(argv)
except ValueError as e:
    if "--model-params" in str(e):
        print(f"usage: --model-params '<json object>' — {e}")
        sys.exit(2)
    raise

When it happens

Trigger: Invoking the CLI with --model-params as the last token, e.g. `dcode --model-params`, or with the flag followed only by whitespace.

Common situations: Copy-pasting a command where the JSON value ended up on a second line that the shell dropped; a script template where the JSON variable was empty; forgetting the value while exploring flags interactively.

Understand the failure class

Background: "Unknown argument", "Invalid value", and "must be one of": invalid CLI argument errors explained — this error's family across 35 libraries.

Related errors

• Unbalanced braces in --model-params value
• enable_interpreter=True is not supported with a remote sandb
• Unclosed {quote} in --model-params value
• Error: Could not read the skill trust store: {exc}
• Error: Invalid agent name: {error_msg}

AI-assisted analysis of langchain-ai/deepagents@a1af029e6e (2026-08-29). Data as JSON: /api/errors/c04efbb5009185a6. Report an issue: GitHub.