fix: allow omitting temperature for models that reject it (e.g. Opus 4.7+)

[yakir-shriker]

Problem

Anthropic Opus 4.7 / 4.8 reject the temperature parameter:

invalid_request_error: `temperature` is deprecated for this model.

HolmesGPT always sends temperature (tool_calling_llm.py → completion(..., temperature=TEMPERATURE), default 1e-8), so these models 400 on every call. drop_params=True doesn't help — the deprecation is an Anthropic runtime behavior, not in LiteLLM's static param metadata (verified with litellm up to 1.86.2), so it isn't stripped.

Fix

Make TEMPERATURE omittable: an empty string / none / null yields None, and completion() already skips the param when it's None — so no temperature is sent. Models that accept temperature are unaffected (default unchanged).

TEMPERATURE=none   # send no temperature param

Verification

• Unit tests added (tests/common/test_temperature_env.py).
• End-to-end via a LiteLLM gateway: with TEMPERATURE=none, Opus 4.6 / 4.7 / 4.8 all complete successfully; without it, 4.7/4.8 return the 400 above.

Signed-off-by: Yakir Shriker yakirshr@gmail.com

Summary by CodeRabbit

• New Features

  • Temperature setting can now be disabled or nullified by setting the environment variable to empty string, "none", or "null" values, providing more flexible configuration control.

• Tests

  • Added tests verifying temperature configuration behavior under various conditions, including disabled states and default values.

[claude]

Claude Code Review

This pull request is from a fork — automated review is disabled. A repository maintainer can comment @claude review to run a one-time review.

[coderabbitai]

Walkthrough

A new _load_temperature() helper function is added to holmes/common/env_vars.py that parses the TEMPERATURE environment variable, returning None when set to empty, "none", or "null" values, and returning the float value otherwise with a default of 0.00000001. The module constant TEMPERATURE now has type Optional[float]. Test coverage verifies all three parsing behaviors.

Changes

Temperature optional parsing

Layer / File(s)	Summary
Temperature parsing implementation holmes/common/env_vars.py	New _load_temperature() function normalizes TEMPERATURE by returning None for empty/disabled strings ("", "none", "null") and converting valid numeric strings to float, with default 0.00000001 when unset. Module constant TEMPERATURE is now Optional[float] and computed via this helper.
Temperature parsing tests tests/common/test_temperature_env.py	Three test functions patch os.environ to verify _load_temperature() returns None for disabled values, returns the parsed float for numeric values like "0.5", and returns the default 0.00000001 when unset.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

Possibly related PRs

• HolmesGPT/holmesgpt#1962: Updates DefaultLLM.completion to handle temperature=None values that result from the new optional temperature parsing by stripping None before forwarding to providers.
• HolmesGPT/holmesgpt#698: Changes how TEMPERATURE is sourced and passed to LLM completion calls to work with the new optional temperature value.

Suggested reviewers

• arikalon1
• naomi-robusta

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately describes the main change: making temperature omittable for models that reject it (e.g., Opus 4.7+), which directly corresponds to the code changes that make TEMPERATURE optional and the feature to interpret specific values as None.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[netlify]

✅ Deploy Preview for holmes-docs ready!

Name	Link
🔨 Latest commit	742500b
🔍 Latest deploy log	https://app.netlify.com/projects/holmes-docs/deploys/6a1c5c59c7e3970008456b79
😎 Deploy Preview	https://deploy-preview-2109--holmes-docs.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify project configuration.

[coderabbitai]

🧹 Nitpick comments (1)

holmes/common/env_vars.py (1)

68-77: ⚡ Quick win

Consider adding error handling for invalid float values.

If TEMPERATURE is set to an invalid numeric string (e.g., "abc"), the float(raw) call will raise a ValueError at startup. While this fails fast and is easy to debug, wrapping it in a try-except with a more descriptive error message could improve the developer experience.

🛡️ Proposed error handling enhancement

 def _load_temperature() -> Optional[float]:
     # Set TEMPERATURE to an empty string / "none" / "null" to send NO temperature at
     # all. Required for models that reject the parameter (e.g. Anthropic Opus 4.7+:
     # "temperature is deprecated for this model"); LiteLLM's drop_params can't strip
     # it because the deprecation isn't in its static param metadata, so it must be
     # omitted at the source.
     raw = os.environ.get("TEMPERATURE", "0.00000001").strip()
     if raw.lower() in ("", "none", "null"):
         return None
-    return float(raw)
+    try:
+        return float(raw)
+    except ValueError as e:
+        raise ValueError(f"Invalid TEMPERATURE value '{raw}': must be a number, 'none', 'null', or empty") from e

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@holmes/common/env_vars.py` around lines 68 - 77, The _load_temperature
function should guard against invalid numeric strings from the TEMPERATURE env
var: wrap the float(raw) conversion in a try/except ValueError and raise a new
ValueError (or use process logging then re-raise) with a clear message that
includes the invalid raw value and that TEMPERATURE must be a float or one of
"", "none", "null"; update the function _load_temperature to perform this
validation around the float(raw) call and surface a descriptive error instead of
the raw ValueError.

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Nitpick comments:
In `@holmes/common/env_vars.py`:
- Around line 68-77: The _load_temperature function should guard against invalid
numeric strings from the TEMPERATURE env var: wrap the float(raw) conversion in
a try/except ValueError and raise a new ValueError (or use process logging then
re-raise) with a clear message that includes the invalid raw value and that
TEMPERATURE must be a float or one of "", "none", "null"; update the function
_load_temperature to perform this validation around the float(raw) call and
surface a descriptive error instead of the raw ValueError.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: 56e822ff-84f0-4619-a32e-788346e1d7d3

📥 Commits

Reviewing files that changed from the base of the PR and between 1bc25d9 and 742500b.

📒 Files selected for processing (2)

• holmes/common/env_vars.py
• tests/common/test_temperature_env.py