Fixes the Gemini integration example in the README

[vishvaRam]

Description

This update fixes several critical issues in the Gemini integration example:

1. Corrected import: Changed from gemini_complete to gemini_model_complete
   (the correct function name per lightrag/llm/gemini.py)

2. Fixed parameter name: Changed 'model' to 'model_name' in gemini_model_complete() call to match the function signature

3. Added llm_model_name to LightRAG initialization: This is required for gemini_model_complete to retrieve the model name from hashing_kv.global_config

4. Updated to latest model: gemini-1.5-flash → gemini-2.0-flash

Without these changes, users get "404 NOT_FOUND" errors as the code defaults to gpt-4o-mini when model_name is not properly configured.

Tested and verified working with Gemini 2.0 Flash API.

Related Issues

Fixes incorrect example code that causes user confusion and setup failures.
This addresses a documentation bug rather than a code functionality issue.

Changes Made

• Changed import: gemini_complete → gemini_model_complete
• Fixed parameter: model= → model_name= in function call
• Added llm_model_name="gemini-2.0-flash" to LightRAG initialization
• Updated model version: gemini-1.5-flash → gemini-2.0-flash

Files Modified

• README.md - Gemini integration example section

Checklist

• Changes tested locally
• Code reviewed (self-reviewed against source)
• Documentation updated (this IS documentation)
• Unit tests added (N/A - documentation only)

Additional Notes

Why These Changes Matter

The current example in README doesn't match the actual library implementation.
Looking at lightrag/llm/gemini.py lines 438-448, gemini_model_complete()
requires the model name to be configured via:

1. hashing_kv.global_config.get("llm_model_name") (checked first)
2. kwargs.pop("model_name", None) (fallback)

Without proper configuration, it defaults to gpt-4o-mini, causing 404 errors.

Testing Environment

• Gemini 2.0 Flash API
• Document size: 53KB
• Successfully extracted: 125 entities, 84 relations
• All query modes (naive, local, global, hybrid) verified working

Reference

Implementation details: lightrag/llm/gemini.py lines 438-448