[GH-ISSUE #13] iterm firecrawl github arxiv mcp server not function well

bsakaguchi commented

2026-10-18 21:10:12 +00:00

Owner

Originally created by @jincai on GitHub (Apr 2, 2025).
Original GitHub issue: https://github.com/open-webui/mcpo/issues/13

github

{
"error": "HTTP error! Status: 500. Message: Internal Server Error"
}

mcp.shared.exceptions.McpError: Invalid input: [{"code":"invalid_type","expected":"number","received":"null","path":["page"],"message":"Expected number, received null"},{"code":"invalid_type","expected":"number","received":"null","path":["perPage"],"message":"Expected number, received null"}]

iterm

[
"1 lines were output after sending the command to the terminal. Read the last 1 lines of terminal contents to orient yourself. Never assume that the command was executed or that it was successful."
]

INFO: 127.0.0.1:56345 - "OPTIONS /iterm/write_to_terminal HTTP/1.1" 200 OK
Calling write_to_terminal with arguments: {'command': 'ls ~/Desktop'}
INFO: 127.0.0.1:56345 - "POST /iterm/write_to_terminal HTTP/1.1" 200 OK

arxiv

[
"Error: Invalid date format - Parser must be a string or character stream, not NoneType"
]

mcp.shared.exceptions.McpError: Invalid input: [{"code":"invalid_type","expected":"number","received":"null","path":["page"],"message":"Expected number, received null"},{"code":"invalid_type","expected":"number","received":"null","path":["perPage"],"message":"Expected number, received null"}]
Calling search_papers with arguments: {'query': 'AI', 'max_results': 10, 'date_from': None, 'date_to': None, 'categories': ['cs.AI', 'cs.LG', 'cs.CL']}
INFO: 127.0.0.1:56772 - "POST /arxiv/search_papers HTTP/1.1" 200 OK

firecrawl

[
"Cannot read properties of undefined (reading 'status')"
]

[info] [2025-04-02T00:03:53.719Z] Received request for tool: firecrawl_scrape
[info] Starting scrape for URL: https://coinmarketcap.com/currencies/bitcoin/ with options: {"formats":["markdown"],"onlyMainContent":true,"includeTags":null,"excludeTags":null,"waitFor":null,"timeout":null,"actions":null,"extract":null,"mobile":null,"skipTlsVerification":null,"removeBase64Images":null,"location":null}
[info] Request completed in 5011ms
INFO: 127.0.0.1:56896 - "POST /firecrawl/firecrawl_scrape HTTP/1.1" 200 OK

Originally created by @jincai on GitHub (Apr 2, 2025). Original GitHub issue: https://github.com/open-webui/mcpo/issues/13 **github** { "error": "HTTP error! Status: 500. Message: Internal Server Error" } mcp.shared.exceptions.McpError: Invalid input: [{"code":"invalid_type","expected":"number","received":"null","path":["page"],"message":"Expected number, received null"},{"code":"invalid_type","expected":"number","received":"null","path":["perPage"],"message":"Expected number, received null"}] **iterm** [ "1 lines were output after sending the command to the terminal. Read the last 1 lines of terminal contents to orient yourself. Never assume that the command was executed or that it was successful." ] INFO: 127.0.0.1:56345 - "OPTIONS /iterm/write_to_terminal HTTP/1.1" 200 OK Calling write_to_terminal with arguments: {'command': 'ls ~/Desktop'} INFO: 127.0.0.1:56345 - "POST /iterm/write_to_terminal HTTP/1.1" 200 OK **arxiv** [ "Error: Invalid date format - Parser must be a string or character stream, not NoneType" ] mcp.shared.exceptions.McpError: Invalid input: [{"code":"invalid_type","expected":"number","received":"null","path":["page"],"message":"Expected number, received null"},{"code":"invalid_type","expected":"number","received":"null","path":["perPage"],"message":"Expected number, received null"}] Calling search_papers with arguments: {'query': 'AI', 'max_results': 10, 'date_from': None, 'date_to': None, 'categories': ['cs.AI', 'cs.LG', 'cs.CL']} INFO: 127.0.0.1:56772 - "POST /arxiv/search_papers HTTP/1.1" 200 OK **firecrawl** [ "Cannot read properties of undefined (reading 'status')" ] [info] [2025-04-02T00:03:53.719Z] Received request for tool: firecrawl_scrape [info] Starting scrape for URL: https://coinmarketcap.com/currencies/bitcoin/ with options: {"formats":["markdown"],"onlyMainContent":true,"includeTags":null,"excludeTags":null,"waitFor":null,"timeout":null,"actions":null,"extract":null,"mobile":null,"skipTlsVerification":null,"removeBase64Images":null,"location":null} [info] Request completed in 5011ms INFO: 127.0.0.1:56896 - "POST /firecrawl/firecrawl_scrape HTTP/1.1" 200 OK

bsakaguchi closed this issue

2026-10-18 21:10:13 +00:00

bsakaguchi commented

2026-10-18 21:10:14 +00:00

Author

Owner

@peanutsugarlove commented on GitHub (Apr 3, 2025):

You should have encountered the problem I encountered because your model did not provide all parameters, resulting in some types not matching and reporting errors.

i ask tool_endpoint_fetch_post What are the parameters
The tool_endpoint_fetch_post function has the following parameters:

url (required)
- Type: string
- Description: The URL to fetch content from.
max_length (optional)
- Type: integer
- Description: The maximum number of characters to return in the response. Useful for limiting the output size.
start_index (optional)
- Type: integer
- Description: The starting character index for the returned content. Useful if you need to fetch additional content after a truncated response.
raw (optional)
- Type: boolean
- Description: If set to true, the function will return the raw HTML content of the page without simplification. Default is false.

Example Usage:

{
  "url": "https://example.com",
  "max_length": 500,
  "start_index": 0,
  "raw": false
}

Because the model marks it as optional

Before solving this problem, it is best to ask the model to force all parameters to be output in the prompt words

 {
      fromPath: z.string().min(1, 'Source path must not be empty'),
      toPath: z.string().min(1, 'Destination path must not be empty'),
      overwrite: z.boolean().optional().default(false)
    },

The above is a definition of tool parameters, and the last parameter may not output

@peanutsugarlove commented on GitHub (Apr 3, 2025): You should have encountered the problem I encountered because your model did not provide all parameters, resulting in some types not matching and reporting errors. i ask tool_endpoint_fetch_post What are the parameters The `tool_endpoint_fetch_post` function has the following parameters: 1. **`url`** (required) - Type: `string` - Description: The URL to fetch content from. 2. **`max_length`** (optional) - Type: `integer` - Description: The maximum number of characters to return in the response. Useful for limiting the output size. 3. **`start_index`** (optional) - Type: `integer` - Description: The starting character index for the returned content. Useful if you need to fetch additional content after a truncated response. 4. **`raw`** (optional) - Type: `boolean` - Description: If set to `true`, the function will return the raw HTML content of the page without simplification. Default is `false`. ### Example Usage: ```json { "url": "https://example.com", "max_length": 500, "start_index": 0, "raw": false } ``` Because the model marks it as optional Before solving this problem, it is best to ask the model to force all parameters to be output in the prompt words ```js { fromPath: z.string().min(1, 'Source path must not be empty'), toPath: z.string().min(1, 'Destination path must not be empty'), overwrite: z.boolean().optional().default(false) }, ``` The above is a definition of tool parameters, and the last parameter may not output

bsakaguchi commented

2026-10-18 21:10:14 +00:00

Author

Owner

@tjbck commented on GitHub (Apr 3, 2025):

Most local models struggle with tool calls, I'd highly recommend using gpt-4o for tool use cases at least for testing purposes.

@tjbck commented on GitHub (Apr 3, 2025): Most local models struggle with tool calls, I'd highly recommend using gpt-4o for tool use cases at least for testing purposes.

bsakaguchi referenced this issue

2026-10-18 21:12:11 +00:00

[PR #11] [CLOSED] Add dockerfile #141

Rows
Columns

[GH-ISSUE #13] iterm firecrawl github arxiv mcp server not function well #11

Example Usage: