Add OpenAI's GPT-5.4 model

[PeterDaveHello]

Rererence:

• https://openai.com/index/introducing-gpt-5-4/
• https://developers.openai.com/api/docs/models/gpt-5.4

[qodo-free-for-open-source-projects]

Review Summary by Qodo

Add OpenAI's GPT-5.4 model support

✨ Enhancement

Walkthroughs

Description

• Add GPT-5.4 model support with 272K token context window
• Register both base and dated model variants in token limits
• Add unit tests for GPT-5.4 token limit validation
• Update GPT-5 model detection tests to include new variants

Diagram

flowchart LR
  A["GPT-5.4 Models"] --> B["Token Limits Config"]
  A --> C["Unit Tests"]
  B --> D["272K Context Window"]
  C --> E["Model Validation"]

Loading

File Changes

1. pr_agent/algo/__init__.py ✨ Enhancement +2/-0

Register GPT-5.4 models with token limits

• Add gpt-5.4 model with 272K token limit
• Add gpt-5.4-2026-03-05 dated variant with 272K token limit
• Include comments explaining 272K as safe default without opt-in 1M context

pr_agent/algo/init.py

---

2. tests/unittest/test_get_max_tokens.py 🧪 Tests +13/-0

Add GPT-5.4 token limit unit tests

• Add parametrized test for both GPT-5.4 model variants
• Verify both gpt-5.4 and gpt-5.4-2026-03-05 return 272000 tokens
• Test with mocked settings to validate token limit retrieval

tests/unittest/test_get_max_tokens.py

---

3. tests/unittest/test_litellm_reasoning_effort.py 🧪 Tests +1/-1

Update GPT-5 model detection tests

• Update GPT-5 model detection test to include new variants
• Add gpt-5.4 and gpt-5.4-2026-03-05 to model version test list
• Ensure reasoning_effort logic applies to new GPT-5.4 models

tests/unittest/test_litellm_reasoning_effort.py

---

[qodo-free-for-open-source-projects]

Code Review by Qodo

🐞 Bugs (0) 📘 Rule violations (3) 📎 Requirement gaps (0)

1. gpt5_models line over 120 📘 Rule violation ✓ Correctness

Description

The modified gpt5_models assignment is a single line longer than the repository's Ruff
line-length = 120, which can cause lint failures. Reformatting this list into a multi-line
structure will restore compliance.

Code

tests/unittest/test_litellm_reasoning_effort.py[310]

+        gpt5_models = ["gpt-5-2025-08-07", "gpt-5.1", "gpt-5.4", "gpt-5.4-2026-03-05", "gpt-5-turbo", "gpt-5.1-codex", "gpt-5.3-codex"]

Evidence

PR Compliance ID 8 requires adherence to Ruff style including 120-character lines; the repo config
sets line-length = 120 and the changed gpt5_models assignment is a long single line that exceeds
this limit.

AGENTS.md
pyproject.toml[47-49]
tests/unittest/test_litellm_reasoning_effort.py[310-310]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
The updated `gpt5_models` assignment exceeds the repository Ruff `line-length = 120` requirement, which may fail lint checks.
## Issue Context
Ruff is configured with `line-length = 120` in `pyproject.toml`.
## Fix Focus Areas
- tests/unittest/test_litellm_reasoning_effort.py[310-310]

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

---

2. MAX_TOKENS hardcodes gpt-5.4 📘 Rule violation ⛯ Reliability ⭐ New

Description

The PR introduces hardcoded runtime token limits for gpt-5.4 models in code instead of using the
repo’s TOML/Dynaconf override mechanisms. This makes runtime behavior harder to change/review and
conflicts with the configuration-driven requirement.

Code

pr_agent/algo/init.py[R45-46]

+    'gpt-5.4': 272000,  # 272K safe default without opt-in 1M context parameters
+    'gpt-5.4-2026-03-05': 272000,  # 272K safe default without opt-in 1M context parameters

Evidence

Compliance requires avoiding newly introduced hardcoded configuration values in application logic
and using .pr_agent.toml or pr_agent/settings/ overrides instead. The added gpt-5.4 entries
hardcode max token values directly in MAX_TOKENS.

AGENTS.md
pr_agent/algo/init.py[45-46]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
New runtime behavior for `gpt-5.4` max tokens is hardcoded in `MAX_TOKENS` rather than being configurable via `.pr_agent.toml` or `pr_agent/settings/*.toml` overrides.

## Issue Context
The repo already supports config-based limits via `config.max_model_tokens` / `config.custom_model_max_tokens`. Add a config-driven mechanism for per-model max tokens (or a model override entry) and have code consult it first.

## Fix Focus Areas
- pr_agent/algo/__init__.py[45-46]
- pr_agent/algo/utils.py[994-1012]
- pr_agent/settings/configuration.toml[29-34]

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

---

3. Duplicated fake_settings construction 📘 Rule violation ✓ Correctness ⭐ New

Description

The new unit test duplicates the fake_settings construction pattern instead of reusing a
helper/fixture, increasing maintenance cost and reducing readability. This violates the expectation
to reduce duplication and clarify control flow in touched code.

Code

tests/unittest/test_get_max_tokens.py[R25-32]

+    @pytest.mark.parametrize("model", ["gpt-5.4", "gpt-5.4-2026-03-05"])
+    def test_gpt54_model_max_tokens(self, monkeypatch, model):
+        fake_settings = type('', (), {
+            'config': type('', (), {
+                'custom_model_max_tokens': 0,
+                'max_model_tokens': 0
+            })()
+        })()

Evidence

The compliance rule requires reducing duplication by extracting repeated logic into
variables/helpers/fixtures. The added test introduces another inline fake_settings = type(...
block duplicating the same setup already used in other tests in this file.

tests/unittest/test_get_max_tokens.py[25-32]
Best Practice: Learned patterns

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
The new test adds another inline `fake_settings` construction block, duplicating the same setup repeated across tests.

## Issue Context
Use a pytest fixture (e.g., `@pytest.fixture` returning `fake_settings`) or a small helper function inside the test module/class to generate settings, and reuse it across tests.

## Fix Focus Areas
- tests/unittest/test_get_max_tokens.py[25-37]

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

---

ⓘ The new review experience is currently in Beta. Learn more

Previous review results

Review updated until commit 58db87d

Results up to commit 1bcefef

🐞 Bugs (0) 📘 Rule violations (1) 📎 Requirement gaps (0)

1. gpt5_models line over 120 📘 Rule violation ✓ Correctness

Description

The modified gpt5_models assignment is a single line longer than the repository's Ruff
line-length = 120, which can cause lint failures. Reformatting this list into a multi-line
structure will restore compliance.

Code

tests/unittest/test_litellm_reasoning_effort.py[310]

+        gpt5_models = ["gpt-5-2025-08-07", "gpt-5.1", "gpt-5.4", "gpt-5.4-2026-03-05", "gpt-5-turbo", "gpt-5.1-codex", "gpt-5.3-codex"]

Evidence

PR Compliance ID 8 requires adherence to Ruff style including 120-character lines; the repo config
sets line-length = 120 and the changed gpt5_models assignment is a long single line that exceeds
this limit.

AGENTS.md
pyproject.toml[47-49]
tests/unittest/test_litellm_reasoning_effort.py[310-310]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
The updated `gpt5_models` assignment exceeds the repository Ruff `line-length = 120` requirement, which may fail lint checks.

## Issue Context
Ruff is configured with `line-length = 120` in `pyproject.toml`.

## Fix Focus Areas
- tests/unittest/test_litellm_reasoning_effort.py[310-310]

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

---

[PeterDaveHello]

@naorpeled this one should be good 👍

[qodo-free-for-open-source-projects]

Persistent review updated to latest commit 58db87d