OpenManus with Gemini 2 million token

[tuapuikia]

I manage to use with Gemini 2.0 with 2 million token. Very useful projects.

[manny1185]

how much you pay?

[jahaskell53]

Gemini 2.0 is free. You can create an API key here.

[marcospaulobb]

@jahaskell53 I'm thinking to use Gemini 1.5 Flash that is cheaper but not free. Do you think that 2.0 is better?

[ThyannSeng]

can you shows the config example without your real api key? it's not working for me and shows 2025-03-10 19:46:03.938 | WARNING | __main__:main:19 - Processing your request... 2025-03-10 19:46:03.938 | INFO | app.agent.base:run:137 - Executing step 1/30 2025-03-10 19:46:05.511 | ERROR | app.llm:ask_tool:256 - Authentication failed. Check API key. 2025-03-10 19:46:06.835 | ERROR | app.llm:ask_tool:256 - Authentication failed. Check API key. 2025-03-10 19:46:08.462 | ERROR | app.llm:ask_tool:256 - Authentication failed. Check API key.

[huypham37]

Can you show which api base url you used for Gemini?

[ThyannSeng]

Can you show which api base url you used for Gemini?

` Global LLM configuration
[llm]
model = "gemini-2.0-flash"
base_url = "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash:generateContent?"
api_key = "My-free-gemini-api-key"
max_tokens = 4096
temperature = 0.0

Optional configuration for specific LLM models
[llm.vision]
model = "gemini-2.0-flash-exp"
base_url = "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash:generateContent?"
api_key = "My-free-gemini-api-key"

`

[pazoff]

Can you show which api base url you used for Gemini?

` Global LLM configuration [llm] model = "gemini-2.0-flash" base_url = "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash:generateContent?" api_key = "My-free-gemini-api-key" max_tokens = 4096 temperature = 0.0

Optional configuration for specific LLM models [llm.vision] model = "gemini-2.0-flash-exp" base_url = "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash:generateContent?" api_key = "My-free-gemini-api-key"

`
I am getting( I have set the key):

2025-03-10 09:01:08.741 | ERROR | app.llm:ask_tool:256 - Authentication failed. Check API key.
2025-03-10 09:01:09.817 | ERROR | app.llm:ask_tool:256 - Authentication failed. Check API key.
2025-03-10 09:01:11.136 | ERROR | app.llm:ask_tool:256 - Authentication failed. Check API key.
2025-03-10 09:01:12.480 | ERROR | app.llm:ask_tool:256 - Authentication failed. Check API key.

[ThyannSeng]

Can you show which api base url you used for Gemini?

` Global LLM configuration [llm] model = "gemini-2.0-flash" base_url = "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash:generateContent?" api_key = "My-free-gemini-api-key" max_tokens = 4096 temperature = 0.0

Optional configuration for specific LLM models [llm.vision] model = "gemini-2.0-flash-exp" base_url = "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash:generateContent?" api_key = "My-free-gemini-api-key"

`

if I use
base_url = "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash:generateContent", it shows 2025-03-10 20:16:58.184 | WARNING | __main__:main:19 - Processing your request... 2025-03-10 20:16:58.184 | INFO | app.agent.base:run:137 - Executing step 1/30 2025-03-10 20:16:59.562 | ERROR | app.llm:ask_tool:260 - API error: Error code: 404 2025-03-10 20:17:00.833 | ERROR | app.llm:ask_tool:260 - API error: Error code: 404. No idea how to make it works

[pazoff]

Can you show which api base url you used for Gemini?

` Global LLM configuration [llm] model = "gemini-2.0-flash" base_url = "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash:generateContent?" api_key = "My-free-gemini-api-key" max_tokens = 4096 temperature = 0.0

Optional configuration for specific LLM models [llm.vision] model = "gemini-2.0-flash-exp" base_url = "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash:generateContent?" api_key = "My-free-gemini-api-key"

`

This Config works for me:

#Global LLM configuration
[llm]
model = "gemini-2.0-flash-exp"
base_url = "https://generativelanguage.googleapis.com/v1beta/openai/"
api_key = "YOUR-API-KEY"
max_tokens = 4096
temperature = 0.0

#Optional configuration for specific LLM models
[llm.vision]
model = "gemini-2.0-flash-exp"
base_url = "https://generativelanguage.googleapis.com/v1beta/openai/"
api_key = "YOUR-API-KEY"

taken from https://ai.google.dev/gemini-api/docs/openai#python

[pazoff]

Can you show which api base url you used for Gemini?

Global LLM configuration [llm] model = "gemini-2.0-flash" base_url = "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash:generateContent?" api_key = "My-free-gemini-api-key" max_tokens = 4096 temperature = 0.0 Optional configuration for specific LLM models [llm.vision] model = "gemini-2.0-flash-exp" base_url = "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash:generateContent?" api_key = "My-free-gemini-api-key"

if I use base_url = "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash:generateContent", it shows 2025-03-10 20:16:58.184 | WARNING | __main__:main:19 - Processing your request... 2025-03-10 20:16:58.184 | INFO | app.agent.base:run:137 - Executing step 1/30 2025-03-10 20:16:59.562 | ERROR | app.llm:ask_tool:260 - API error: Error code: 404 2025-03-10 20:17:00.833 | ERROR | app.llm:ask_tool:260 - API error: Error code: 404. No idea how to make it works

Look Here: https://ai.google.dev/gemini-api/docs/openai#python

[ThyannSeng]

Can you show which api base url you used for Gemini?

Global LLM configuration [llm] model = "gemini-2.0-flash" base_url = "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash:generateContent?" api_key = "My-free-gemini-api-key" max_tokens = 4096 temperature = 0.0 Optional configuration for specific LLM models [llm.vision] model = "gemini-2.0-flash-exp" base_url = "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash:generateContent?" api_key = "My-free-gemini-api-key"

This Config works for me:

#Global LLM configuration
[llm]
model = "gemini-2.0-flash-exp"
base_url = "https://generativelanguage.googleapis.com/v1beta/openai/"
api_key = "YOUR-API-KEY"
max_tokens = 4096
temperature = 0.0

#Optional configuration for specific LLM models
[llm.vision]
model = "gemini-2.0-flash-exp"
base_url = "https://generativelanguage.googleapis.com/v1beta/openai/"
api_key = "YOUR-API-KEY"

taken from https://ai.google.dev/gemini-api/docs/openai#python

Thanks! It's working.

[pazoff]

Can you show which api base url you used for Gemini?

Global LLM configuration [llm] model = "gemini-2.0-flash" base_url = "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash:generateContent?" api_key = "My-free-gemini-api-key" max_tokens = 4096 temperature = 0.0 Optional configuration for specific LLM models [llm.vision] model = "gemini-2.0-flash-exp" base_url = "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash:generateContent?" api_key = "My-free-gemini-api-key"

This Config works for me:

#Global LLM configuration
[llm]
model = "gemini-2.0-flash-exp"
base_url = "https://generativelanguage.googleapis.com/v1beta/openai/"
api_key = "YOUR-API-KEY"
max_tokens = 4096
temperature = 0.0

#Optional configuration for specific LLM models
[llm.vision]
model = "gemini-2.0-flash-exp"
base_url = "https://generativelanguage.googleapis.com/v1beta/openai/"
api_key = "YOUR-API-KEY"

taken from https://ai.google.dev/gemini-api/docs/openai#python

Thanks! It's working.

But there is other problems:
| ERROR | app.llm:ask_tool:252 - Validation error in ask_tool: Message must contain either 'content' or 'tool_calls'

[manny1185]

Gemini 2.0 is free. You can create an API key here.

Thanks,i have supply one

[krigeta]

can we use thinking models as well?

[marko999]

thank you for the config, that work, but that leads to rate limiting errors, now this is free 1mil token api, so it cannot be on server side and from the front side I see only one or more actual request made, so something is off in that setup

[yd-ccccai]

Using gemini api will end up with: Validation error in ask_tool: Message must contain either 'content' or 'tool_calls'
Either flash or flash-exp.
Maybe gemini has something error with function_call

[kbrajwani]

Using gemini api will end up with: Validation error in ask_tool: Message must contain either 'content' or 'tool_calls' Either flash or flash-exp. Maybe gemini has something error with function_call

You can easily bypass this by editing app/agent/toolcall.py.

Add this call into a file

class Response:
    def __init__(self):
        self.tool_calls = None
        self.content = None

then add a while clause before calling the LLM.

while (response.tool_calls == None) and (response.content == None):
            
            response = await self.llm.ask_tool(
                messages=self.messages,
                system_msgs=[Message.system_message(self.system_prompt)]
                if self.system_prompt
                else None,
                tools=self.available_tools.to_params(),
                tool_choice=self.tool_choices,
            )

It will try to call LLM until it provides any action to take. Sometimes, LLM is not able to provide the next steps, so we end up with the issue.

[DKauser]

Using gemini api will end up with: Validation error in ask_tool: Message must contain either 'content' or 'tool_calls' Either flash or flash-exp. Maybe gemini has something error with function_call

You can easily bypass this by editing app/agent/toolcall.py.

Add this call into a file

class Response:
    def __init__(self):
        self.tool_calls = None
        self.content = None

then add a while clause before calling the LLM.

while (response.tool_calls == None) and (response.content == None):
            
            response = await self.llm.ask_tool(
                messages=self.messages,
                system_msgs=[Message.system_message(self.system_prompt)]
                if self.system_prompt
                else None,
                tools=self.available_tools.to_params(),
                tool_choice=self.tool_choices,
            )

It will try to call LLM until it provides any action to take. Sometimes, LLM is not able to provide the next steps, so we end up with the issue.

Would this resolve an issue with "Error in PlanningFlow: Parameter 'title' is required for command: create"?

[PBritto697]

If gemini allows millions of tokens, why keep max_tokens = 4096? Wasn't it supposed to be max_token= 1056768?