Skip to content

lasagnastack.llm

lasagnastack.llm

make_client 

make_client(
    model: str | None = None,
    *,
    reasoning_max_tokens: int = 4000,
    reasoning_effort: str | None = None,
    total_max_tokens: int | None = None,
) -> LLMClient

Instantiate the correct LLMClient for the given model string.

Dispatches on the provider prefix (the first path segment of the model name). The prefix is stripped before the model name is forwarded to the client constructor.

Supported prefixes:

  • gemini/ → :class:~lasagnastack.llm.gemini.GeminiClient
  • openrouter/ → :class:~lasagnastack.llm.openrouter.OpenRouterClient

When model is None, the LSNSTK_LLM_MODEL environment variable is read, falling back to "gemini/gemini-2.5-flash".

Parameters:

Name Type Description Default
model str | None

Full model string including provider prefix, e.g. "gemini/gemini-2.5-flash" or "openrouter/deepseek/deepseek-v3.2".

 None
reasoning_max_tokens int

Token budget for internal reasoning. Forwarded as thinking_budget to GeminiClient and as reasoning.max_tokens to OpenRouterClient (when no effort level is set).

 4000
reasoning_effort str | None

Qualitative effort level ("none" / "minimal" / "low" / "medium" / "high" / "xhigh"). Forwarded to OpenRouterClient where it takes priority over reasoning_max_tokens; accepted but ignored by GeminiClient.

 None
total_max_tokens int | None

Ceiling on all output tokens combined (reasoning + visible response). Forwarded to OpenRouterClient as the max_tokens API parameter; accepted but ignored by GeminiClient.

 None

Returns:

Type Description
LLMClient

A concrete LLMClient instance for the resolved provider.

Raises:

Type Description
ValueError

If the provider prefix is not recognised.
Source code in src/lasagnastack/llm/__init__.py 
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
def make_client(
    model: str | None = None,
    *,
    reasoning_max_tokens: int = 4000,
    reasoning_effort: str | None = None,
    total_max_tokens: int | None = None,
) -> LLMClient:
    """Instantiate the correct ``LLMClient`` for the given model string.

    Dispatches on the provider prefix (the first path segment of the model
    name). The prefix is stripped before the model name is forwarded to the
    client constructor.

    Supported prefixes:

    * ``gemini/`` → :class:`~lasagnastack.llm.gemini.GeminiClient`
    * ``openrouter/`` → :class:`~lasagnastack.llm.openrouter.OpenRouterClient`

    When ``model`` is ``None``, the ``LSNSTK_LLM_MODEL`` environment variable
    is read, falling back to ``"gemini/gemini-2.5-flash"``.

    Args:
        model: Full model string including provider prefix, e.g.
            ``"gemini/gemini-2.5-flash"`` or
            ``"openrouter/deepseek/deepseek-v3.2"``.
        reasoning_max_tokens: Token budget for internal reasoning. Forwarded
            as ``thinking_budget`` to ``GeminiClient`` and as
            ``reasoning.max_tokens`` to ``OpenRouterClient`` (when no effort
            level is set).
        reasoning_effort: Qualitative effort level (``"none"`` /
            ``"minimal"`` / ``"low"`` / ``"medium"`` / ``"high"`` /
            ``"xhigh"``). Forwarded to ``OpenRouterClient`` where it takes
            priority over ``reasoning_max_tokens``; accepted but ignored by
            ``GeminiClient``.
        total_max_tokens: Ceiling on all output tokens combined (reasoning +
            visible response). Forwarded to ``OpenRouterClient`` as the
            ``max_tokens`` API parameter; accepted but ignored by
            ``GeminiClient``.

    Returns:
        A concrete ``LLMClient`` instance for the resolved provider.

    Raises:
        ValueError: If the provider prefix is not recognised.
    """
    resolved = model or os.getenv("LSNSTK_LLM_MODEL", "gemini/gemini-2.5-flash")
    prefix = resolved.split("/")[0]

    if prefix == "gemini":
        from lasagnastack.llm.gemini import GeminiClient

        return GeminiClient(
            model=resolved,
            reasoning_max_tokens=reasoning_max_tokens,
            reasoning_effort=reasoning_effort,
            total_max_tokens=total_max_tokens,
        )

    if prefix == "openrouter":
        from lasagnastack.llm.openrouter import OpenRouterClient

        return OpenRouterClient(
            model=resolved,
            reasoning_max_tokens=reasoning_max_tokens,
            reasoning_effort=reasoning_effort,
            total_max_tokens=total_max_tokens,
        )

    raise ValueError(
        f"Unknown LLM provider prefix {prefix!r} in model {resolved!r}. "
        "Expected 'gemini' or 'openrouter'."
    )