BerriAI/litellm · error · AnthropicContextManagementError

context_management.compact_20260112.trigger.value must be at

Error message

context_management.compact_20260112.trigger.value must be at least {COMPACT_MIN_TRIGGER_TOKENS} tokens

What it means

Raised by the in-gateway context_management polyfill when validating the compact_20260112 edit specification. The trigger threshold ('context_management.compact_20260112.trigger.value') must be an int of at least 50,000 tokens (COMPACT_MIN_TRIGGER_TOKENS); smaller values are rejected outright rather than clamped, because too-low thresholds would fire compaction constantly. Non-int values get a warning and fall back to the 150,000 default instead.

Source

Thrown at litellm/llms/anthropic/experimental_pass_through/context_management/editors/compact.py:562

    """
    warnings: Final[list[str]] = []
    trigger: Final = edit_spec.get("trigger") or {}
    if not isinstance(trigger, dict):
        warnings.append("trigger_not_a_dict_using_default")
        return COMPACT_DEFAULT_TRIGGER_TOKENS, warnings

    trigger_type: Final = trigger.get("type", "input_tokens")
    if trigger_type != "input_tokens":
        warnings.append(f"unsupported_trigger_type_{trigger_type}_using_input_tokens")

    value: Final = trigger.get("value")
    if value is None:
        return COMPACT_DEFAULT_TRIGGER_TOKENS, warnings
    if not isinstance(value, int):
        warnings.append("trigger_value_not_int_using_default")
        return COMPACT_DEFAULT_TRIGGER_TOKENS, warnings
    if value < COMPACT_MIN_TRIGGER_TOKENS:
        raise AnthropicContextManagementError(
            status_code=400,
            message=(
                f"context_management.compact_20260112.trigger.value must be at "
                f"least {COMPACT_MIN_TRIGGER_TOKENS} tokens"
            ),
        )
    return value, warnings


def _build_summary_prompt(edit_spec: Mapping[str, object], tools: Sequence[Mapping[str, object]] | None) -> str:
    custom: Final = edit_spec.get("instructions")
    if isinstance(custom, str) and custom.strip():
        return custom
    prompt = COMPACT_DEFAULT_INSTRUCTIONS
    if tools:
        prompt = f"{prompt}{COMPACT_NO_TOOL_CALLS_SUFFIX}"
    return prompt

View on GitHub (pinned to 6c2dcb801b)

Solutions

1. Set trigger.value to an int >= 50,000, e.g. 50_000 is the minimum accepted.
2. For local testing of compaction, you cannot lower the threshold; instead send a long conversation that exceeds it.
3. Omit trigger entirely to accept the 150,000 default.
4. Keep the value a plain int (not a float or string) or the default is silently used with a warning.

Example fix

# before
context_management = {
    "edits": [{"type": "compact_20260112", "trigger": {"type": "input_tokens", "value": 5000}}]
}

# after
context_management = {
    "edits": [{"type": "compact_20260112", "trigger": {"type": "input_tokens", "value": 50000}}]
}

COMPACT_MIN_TRIGGER_TOKENS = 50_000

def validate_compact_trigger(trigger: dict) -> int:
    value = trigger.get("value", 150_000)
    if not isinstance(value, int) or isinstance(value, bool):
        return 150_000  # gateway will warn and use default
    if value < COMPACT_MIN_TRIGGER_TOKENS:
        raise ValueError(f"trigger.value must be >= {COMPACT_MIN_TRIGGER_TOKENS}")
    return value

def is_valid_compact_trigger(trigger: object) -> bool:
    if not isinstance(trigger, dict):
        return False
    v = trigger.get("value")
    return v is None or (isinstance(v, int) and not isinstance(v, bool) and v >= 50_000)

try:
    resp = litellm.anthropic_messages(**body)
except Exception as e:
    if "must be at least" in str(e) and "trigger.value" in str(e):
        body["context_management"]["edits"][0]["trigger"]["value"] = 50_000
        resp = litellm.anthropic_messages(**body)
    else:
        raise

When it happens

Trigger: Sending an Anthropic /v1/messages request with context_management={"edits": [{"type": "compact_20260112", "trigger": {"value": 10000}}]} — any int below 50,000. Note float values like 50000.0 hit the 'not int' warning path instead and use the default.

Common situations: Developers testing compaction with small token counts to force it to trigger; porting configs from providers with lower compaction thresholds; assuming the value is a percentage or a ratio rather than an absolute token count.

Related errors

• WebSearchInterception: missing follow-up messages
• Invalid first message. Should always start with 'role'='user
• Unable to parse anthropic tool result for message: {message}
• Unable to parse anthropic file message: {message}
• Either file_data or file_id must be present in the file mess

AI-assisted analysis of BerriAI/litellm@6c2dcb801b (2026-08-15). Data as JSON: /api/errors/d50db9ec8261be53. Report an issue: GitHub.