fix(backend): replace decommissioned gemma2-9b-it with llama-3.3-70b-versatile

[Prabhatyadav60]

Fixes #134

Description

The current hardcoded model gemma2-9b-it has been decommissioned by Groq, causing the backend to crash with groq.BadRequestError immediately upon running any analysis (Bias, Chat, or Fact-Check).

Changes Proposed

• Updated all backend instances of gemma2-9b-it to the stable llama-3.3-70b-versatile model.
• Verified compatibility with existing prompts.

Verification

• ✅ Bias Analysis: Success (Score generated).
• ✅ Chat: Success (Responds to context).

Reasoning

• Compatibility: llama-3.3-70b is natively supported by the existing Groq client, requiring zero code refactoring.
• Capabilities: As a 70B parameter model, it offers superior reasoning for Bias and Fact-Check analysis compared to the decommissioned 9B model.
• Stability: It is the current stable flagship on Groq, ensuring long-term reliability.

Summary by CodeRabbit

• Chores
  • Updated the underlying AI model used across core processing modules to improve response quality and consistency.
  • No user-facing behavior or workflows were changed; existing prompts, outputs, and error handling remain the same.

[coderabbitai]

📝 Walkthrough

Walkthrough

This PR updates the LLM model identifier from "gemma2-9b-it" to "llama-3.3-70b-versatile" across five backend modules. No other logic, control flow, or error handling is modified.

Changes

Cohort / File(s)	Summary
Bias Detection backend/app/modules/bias_detection/check_bias.py	Replaced Groq model parameter: gemma2-9b-it → llama-3.3-70b-versatile in the API call.
Chat backend/app/modules/chat/llm_processing.py	Updated LLM invocation model string from gemma2-9b-it to llama-3.3-70b-versatile.
Facts Check backend/app/modules/facts_check/llm_processing.py	Two chat completion calls changed to use llama-3.3-70b-versatile instead of gemma2-9b-it.
Langgraph Nodes backend/app/modules/langgraph_nodes/judge.py, backend/app/modules/langgraph_nodes/sentiment.py	Replaced module-level/inline Groq model initializations with llama-3.3-70b-versatile (was gemma2-9b-it).

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~3 minutes

Possibly related PRs

• Added Sentiment Analysis Node using Groq SDK #101: Updates the Groq model name to "llama-3.3-70b-versatile" in sentiment/judge modules and other LLM calls; directly related to the same model-string replacement.

Poem

🐰 A tiny hop through files I tread,
Swapping names where LLMs are fed,
gemma bowed out, llama took the stage,
Five small edits on the code-stage page,
Nose a-twitch — the build moves ahead.

🚥 Pre-merge checks | ✅ 2 | ❌ 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 (2 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately describes the primary change: replacing a decommissioned model with a new one across the backend.
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.

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

✨ Finishing touches

• 📝 Generate docstrings

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment

---

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.}

[coderabbitai]

Actionable comments posted: 0

Caution

Some comments are outside the diff and can’t be posted inline due to platform limitations.

⚠️ Outside diff range comments (1)

backend/app/modules/facts_check/llm_processing.py (1)

143-148: ⚠️ Potential issue | 🟡 Minor

Pre-existing bug: parsed is used unbound if JSON parsing fails.

Not introduced by this PR, but worth noting: if json.loads raises on line 144, the except block logs the error but doesn't set parsed or continue/skip. Line 148 then references parsed, which will raise NameError and crash the entire function.

This risk increases slightly with a model change, as the new model's output format may differ.

Suggested fix

             try:
                 parsed = json.loads(content)
             except Exception as parse_err:
                 logger.error(f"LLM JSON parse error: {parse_err}")
+                parsed = {
+                    "verdict": "Unknown",
+                    "explanation": f"Parse error: {parse_err}",
+                    "original_claim": claim,
+                    "source_link": source,
+                }

🧹 Nitpick comments (1)

backend/app/modules/bias_detection/check_bias.py (1)

64-64: Consider centralizing the model identifier into a shared constant.

The string "llama-3.3-70b-versatile" is now hardcoded in five separate files. The next time the model needs to change (as happened with this PR), every occurrence must be found and updated. A single constant (e.g., GROQ_MODEL in a shared config module or environment variable) would make future swaps a one-line change.

Example

In a shared config (e.g., backend/app/config.py):

GROQ_MODEL = os.getenv("GROQ_MODEL", "llama-3.3-70b-versatile")

Then in each module:

-            model="llama-3.3-70b-versatile",
+            model=GROQ_MODEL,