add utility to get installed libraries

[KRRT7]

User description

this is for future use

I'm not going to integrate it with the full context yet because our current implementation expects objects to be the same when comparing, this is a problem because we can use a library like polars which is faster than pandas in a few contexts and still be "correct" but we will not accept it due to the types of the objects

---

PR Type

Enhancement

---

Description

• Add utility to list installed Python packages

• Filter out blacklisted development packages

• Integrate installed_packages into AI POST payloads

• Add type hints to contextmanager functions

---

Changes walkthrough 📝

Relevant files

Enhancement

aiservice.py Add installed packages support to AI client                            codeflash/api/aiservice.py Imported get_installed_packages into client Stored installed packages on client initialization Added installed_packages to POST payload +4/-0      code_utils.py Implement installed packages retrieval utility                      codeflash/code_utils/code_utils.py Imported importlib.metadata and TYPE_CHECKING Defined blacklist_installed_pkgs set Added get_installed_packages function Annotated contextmanager functions with Generators +34/-2

---

Need help?

Type /help how to ... in the comments thread for any questions about PR-Agent usage.

Check out the documentation for more information.

[github-actions]

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

⏱️ Estimated effort to review: 2 🔵🔵⚪⚪⚪

🧪 No relevant tests

🔒 No security concerns identified

⚡ Recommended focus areas for review Payload Initialization The code assumes payload is a dict when adding installed_packages. If payload is None, this will raise an exception. Consider initializing payload to an empty dict or guarding this assignment. url = f"{self.base_url}/ai{endpoint}" if method.upper() == "POST": if self.installed_packages: payload["installed_packages"] = self.installed_packages json_payload = json.dumps(payload, indent=None, default=pydantic_encoder) Order Determinism get_installed_packages returns package names in arbitrary order. You may want to sort the list before returning for deterministic payloads and easier testing. def get_installed_packages() -> list[str]: pkgs = importlib.metadata.packages_distributions().keys() return [ pkg for pkg in pkgs if not any(blacklisted in pkg for blacklisted in blacklist_installed_pkgs) and not pkg.startswith("_") ] Missing Version Info The utility only returns package names without versions, which may not be sufficient for reproducibility. Consider including package versions alongside names. def get_installed_packages() -> list[str]: pkgs = importlib.metadata.packages_distributions().keys() return [ pkg for pkg in pkgs if not any(blacklisted in pkg for blacklisted in blacklist_installed_pkgs) and not pkg.startswith("_") ]

[github-actions]

PR Code Suggestions ✨

Explore these optional code suggestions:

Category	Suggestion	Impact
Possible issue	Initialize payload if missing Ensure payload is always a mutable dict before inserting installed packages to avoid a TypeError when payload is None. Initialize or default payload to {} at the start of the POST branch. codeflash/api/aiservice.py [71-72] +if payload is None: + payload = {} if self.installed_packages: payload["installed_packages"] = self.installed_packages Suggestion importance[1-10]: 7 __ Why: Checking and initializing payload prevents a TypeError when it’s None, ensuring the POST branch never fails at runtime.	Medium
	Prevent init failure on metadata error Wrap the call to get_installed_packages() in a try/except block so that any exceptions during metadata retrieval don’t prevent AiServiceClient from initializing. On failure, default to an empty list. codeflash/api/aiservice.py [31] -self.installed_packages = get_installed_packages() +try: + self.installed_packages = get_installed_packages() +except Exception: + self.installed_packages = [] Suggestion importance[1-10]: 5 __ Why: Wrapping get_installed_packages() guards against metadata retrieval errors, improving resilience without being critical.	Low
General	Add package versions Include version information for each package by iterating over importlib.metadata.distributions() and formatting entries as "name==version", which provides more precise context in the payload. codeflash/code_utils/code_utils.py [248-254] def get_installed_packages() -> list[str]: - pkgs = importlib.metadata.packages_distributions().keys() - return [ - pkg - for pkg in pkgs - if not any(blacklisted in pkg for blacklisted in blacklist_installed_pkgs) and not pkg.startswith("_") - ] + pkgs = [] + for dist in importlib.metadata.distributions(): + name = dist.metadata["Name"] + version = dist.version + if not any(bl in name for bl in blacklist_installed_pkgs) and not name.startswith("_"): + pkgs.append(f"{name}=={version}") + return pkgs Suggestion importance[1-10]: 5 __ Why: Including "name==version" entries gives more precise context, but may require downstream adjustments to handle the new format.	Low
	Limit packages in payload Restrict the length of installed_packages added to the payload (e.g., take only the first 100 entries) to avoid creating oversized HTTP requests. codeflash/api/aiservice.py [71-72] if self.installed_packages: - payload["installed_packages"] = self.installed_packages + payload["installed_packages"] = self.installed_packages[:100] Suggestion importance[1-10]: 4 __ Why: Truncating installed_packages to 100 entries can prevent oversized requests, but it’s a design choice with limited urgency.	Low