add support for Gemini 2.5 Flash model and update README

Signed-off-by: JR <[email protected]>

Author: RayChang987

README.md

@@ -44,6 +44,19 @@ If you have a resource-constrained PC, try increasing `HEALTHCHECK_START_PERIOD`
 enough before healthcheck begins.
 For more information, please refer to this [link](https://docs.docker.com/reference/compose-file/services/#healthcheck)
 
+#### Supported Gemini Models
+
+You can specify the Gemini model version using the environment variable `GOOGLE_GEMINI` in your `.env` file.
+The following models are supported:
+
+| Environment Value | Model Name          | Description |
+|------------------|------------------|-------------|
+| `1_pro`          | `gemini-pro`     | Legacy Gemini 1 Pro model (Vertex AI / Generative AI Studio). |
+| `1.5_flash`      | `gemini-1.5-flash` | Lightweight, faster model suitable for low-latency tasks. |
+| `1.5_pro`        | `gemini-1.5-pro` | More capable model for complex reasoning and higher-quality outputs. |
+| `2.5_flash`      | `gemini-2.5-flash` | Latest generation, faster and more accurate than 1.5_flash. |
+
+Set the model by updating your `.env` file:
 ```bash
 cd backend
 cp .env.example .env
@@ -60,9 +73,9 @@ make docker-down
 
 ### Prerequisites
 
-- [`uv`](https://docs.astral.sh/uv/) (for managing Python, virtual environments, and dependencies)  
-- `wget`  
-- `pandoc`  
+- [`uv`](https://docs.astral.sh/uv/) (for managing Python, virtual environments, and dependencies)
+- `wget`
+- `pandoc`
 - `git`
 
 **Step 1**: Install the required dependencies.
@@ -141,12 +154,12 @@ flowchart LR
     id1([Vectorstore]) --- id3([MMR Retriever])
     id1([Vectorstore]) --- id4([BM25 Retriever])
 
-    id2([Semantic Retriever]) -- Retrieved Docs ---> id5([Reranking]) 
+    id2([Semantic Retriever]) -- Retrieved Docs ---> id5([Reranking])
     id3([MMR Retriever]) -- Retrieved Docs ---> id5([Reranking])
     id4([BM25 Retriever]) -- Retrieved Docs ---> id5([Reranking])
 
     id5([Reranking]) ---> id6(top-n docs)
- 
+
 ```
 
 Depending on the input query, each query can be forwarded to any one of the following retrievers,

backend/src/api/routers/chains.py

@@ -60,6 +60,8 @@
         llm = ChatVertexAI(model_name="gemini-1.5-flash", temperature=llm_temp)
     elif os.getenv("GOOGLE_GEMINI") == "1.5_pro":
         llm = ChatVertexAI(model_name="gemini-1.5-pro", temperature=llm_temp)
+    elif os.getenv("GOOGLE_GEMINI") == "2.5_flash":
+        llm = ChatVertexAI(model_name="gemini-2.5-flash", temperature=llm_temp)
     else:
         raise ValueError("GOOGLE_GEMINI environment variable not set to a valid value.")
 

backend/src/api/routers/graphs.py

@@ -81,6 +81,8 @@
         llm = ChatVertexAI(model_name="gemini-1.5-flash", temperature=llm_temp)
     elif os.getenv("GOOGLE_GEMINI") == "1.5_pro":
         llm = ChatVertexAI(model_name="gemini-1.5-pro", temperature=llm_temp)
+    elif os.getenv("GOOGLE_GEMINI") == "2.5_flash":
+        llm = ChatVertexAI(model_name="gemini-2.5-flash", temperature=llm_temp)
     else:
         raise ValueError("GOOGLE_GEMINI environment variable not set to a valid value.")
 
@@ -205,7 +207,7 @@ def parse_agent_output(output: list) -> tuple[str, list[ContextSource], list[str
     embeddings_config=embeddings_config,
     reranking_model_name=hf_reranker,
     use_cuda=use_cuda,
-    inbuilt_tool_calling=True,
+    inbuilt_tool_calling=False,
     fast_mode=fast_mode,
     debug=debug,
     enable_mcp=enable_mcp,