klbr/khoj
1
0
Fork 0
mirror of https://github.com/khoaliber/khoj.git synced 2026-03-02 21:19:12 +00:00
Code Issues Packages Projects Releases Wiki Activity

Support Google Gemini 3

Browse Source 
- Use thinking level for gemini 3 models instead of thinking budget.
- Bump google gemini library
- Add default context, pricing
This commit is contained in:
Debanjum
2025-11-16 23:25:48 -08:00
parent a30c5f245d
commit 57d6ebb1b8
5 changed files with 22 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
src/khoj/processor/conversation/google/utils.py
View File

@@ -203,8 +203,11 @@ def gemini_completion_with_backoff(
response_schema = clean_response_schema(model_kwargs["response_schema"])
thinking_config = None
if deepthought and is_reasoning_model(model_name):
if deepthought and model_name.startswith("gemini-2.5"):
thinking_config = gtypes.ThinkingConfig(thinking_budget=MAX_REASONING_TOKENS_GEMINI, include_thoughts=True)
elif model_name.startswith("gemini-3"):
thinking_level = gtypes.ThinkingLevel.HIGH if deepthought else gtypes.ThinkingLevel.LOW
thinking_config = gtypes.ThinkingConfig(thinking_level=thinking_level, include_thoughts=True)
max_output_tokens = MAX_OUTPUT_TOKENS_FOR_STANDARD_GEMINI
if is_reasoning_model(model_name):
@@ -321,8 +324,11 @@ async def gemini_chat_completion_with_backoff(
formatted_messages, system_instruction = format_messages_for_gemini(messages, system_prompt)
thinking_config = None
if deepthought and is_reasoning_model(model_name):
if deepthought and model_name.startswith("gemini-2.5"):
thinking_config = gtypes.ThinkingConfig(thinking_budget=MAX_REASONING_TOKENS_GEMINI, include_thoughts=True)
elif model_name.startswith("gemini-3"):
thinking_level = gtypes.ThinkingLevel.HIGH if deepthought else gtypes.ThinkingLevel.LOW
thinking_config = gtypes.ThinkingConfig(thinking_level=thinking_level, include_thoughts=True)
max_output_tokens = MAX_OUTPUT_TOKENS_FOR_STANDARD_GEMINI
if is_reasoning_model(model_name):
@@ -563,7 +569,7 @@ def is_reasoning_model(model_name: str) -> bool:
"""
Check if the model is a reasoning model.
"""
return model_name.startswith("gemini-2.5")
return model_name.startswith("gemini-2.5") or model_name.startswith("gemini-3")
def to_gemini_tools(tools: List[ToolDefinition]) -> List[gtypes.ToolDict] | None:

1
src/khoj/processor/conversation/utils.py
View File

@@ -70,6 +70,7 @@ model_to_prompt_size = {
"gpt-5-mini-2025-08-07": 120000,
"gpt-5-nano-2025-08-07": 120000,
# Google Models
"gemini-3-pro-preview": 120000,
"gemini-2.5-flash": 120000,
"gemini-2.5-flash-lite": 120000,
"gemini-2.5-pro": 60000,

5
src/khoj/utils/constants.py
View File

@@ -51,8 +51,9 @@ model_to_cost: Dict[str, Dict[str, float]] = {
"gemini-2.0-flash": {"input": 0.10, "output": 0.40},
"gemini-2.0-flash-lite": {"input": 0.0075, "output": 0.30},
"gemini-2.5-flash-lite": {"input": 0.10, "output": 0.40},
"gemini-2.5-flash": {"input": 0.30, "cache_read_tokens": 0.075, "output": 2.50},
"gemini-2.5-pro": {"input": 1.25, "cache_read_tokens": 0.31, "output": 10.0},
"gemini-2.5-flash": {"input": 0.30, "cache_read_tokens": 0.03, "output": 2.50},
"gemini-2.5-pro": {"input": 1.25, "cache_read_tokens": 0.125, "output": 10.0},
"gemini-3-pro-preview": {"input": 2.00, "cache_read_tokens": 0.20, "output": 12.0},
# Anthropic Pricing: https://www.anthropic.com/pricing#anthropic-api
"claude-3-5-haiku-20241022": {"input": 1.0, "output": 5.0, "cache_read": 0.08, "cache_write": 1.0},
"claude-3-5-haiku@20241022": {"input": 1.0, "output": 5.0, "cache_read": 0.08, "cache_write": 1.0},