feat: add deepseek

[wChenonly]

hi @mattzcarey ，I am having some problems with deepseek.
I get the response data, but I get an error when I use parseJson.

DEBUG Unparsed JSON json { "fileName": "/Users/wchenonly/work/gitHub/f/code-review-gpt/src/common/model/AIModel.ts", "riskScore": 2, "feedbackDetails": "1. The `temperature` parameter in `createDeepSeekModel` is being destructured with a default value of `0.0`, but this default value is not documented or explained. This could lead to confusion or unintended behavior if the default is not appropriate for all use cases. Consider adding a comment or documentation to clarify the default value.\n\n2. The `callStructuredModel` method in `createDeepSeekModel` does not handle potential errors during JSON parsing. If the API response is malformed, this could lead to runtime errors. Consider wrapping the `parseJson` call in a try-catch block to handle such cases gracefully.\n\ntypescript\ncallStructuredModel: async (prompt: string): Promise<IFeedback[]> => {\n try {\n const completion = await client.chat.completions.create({\n messages: [{ role: 'user', content: prompt }],\n model,\n temperature,\n });\n const content = completion.choices[0].message.content;\n return parseJson(content as string);\n } catch (error) {\n logger.error('Failed to parse JSON response', error);\n throw new Error('Failed to parse JSON response');\n }\n},\n```"
}

DEBUG   Escaped JSON \\`\\`\\`\bjson\b\\bn\b{\\bn\b  \"\bfileName\b\": \"/\bUsers\b/\bwchenonly\b/\bwork\b/\bgitHub\b/\bf\b/\bcode\b-\breview\b-\bgpt\b/\bsrc\b/\bcommon\b/\bmodel\b/\bAIModel\b.\bts\b\",\\bn\b  \"\briskScore\b\": \b2\b,\\bn\b  \"\bfeedbackDetails\b\": \"\b1\b. \bThe\b \`\btemperature\b\` \bparameter\b \bin\b \`\bcreateDeepSeekModel\b\` \bis\b \bbeing\b \bdestructured\b \bwith\b \ba\b \bdefault\b \bvalue\b \bof\b \`\b0\b.\b0\b\`, \bbut\b \bthis\b \bdefault\b \bvalue\b \bis\b \bnot\b \bdocumented\b \bor\b \bexplained\b. \bThis\b \bcould\b \blead\b \bto\b \bconfusion\b \bor\b \bunintended\b \bbehavior\b \bif\b \bthe\b \bdefault\b \bis\b \bnot\b \bappropriate\b \bfor\b \ball\b \buse\b \bcases\b. \bConsider\b \badding\b \ba\b \bcomment\b \bor\b \bdocumentation\b \bto\b \bclarify\b \bthe\b \bdefault\b \bvalue\b.\\\bn\b\\\bn2\b. \bThe\b \`\bcallStructuredModel\b\` \bmethod\b \bin\b \`\bcreateDeepSeekModel\b\` \bdoes\b \bnot\b \bhandle\b \bpotential\b \berrors\b \bduring\b \bJSON\b \bparsing\b. \bIf\b \bthe\b \bAPI\b \bresponse\b \bis\b \bmalformed\b, \bthis\b \bcould\b \blead\b \bto\b \bruntime\b \berrors\b. \bConsider\b \bwrapping\b \bthe\b \`\bparseJson\b\` \bcall\b \bin\b \ba\b \btry\b-\bcatch\b \bblock\b \bto\b \bhandle\b \bsuch\b \bcases\b \bgracefully\b.\\\bn\b\\\bn\b\\`\\`\\`\btypescript\b\\\bncallStructuredModel\b: \basync\b (\bprompt\b: \bstring\b): \bPromise\b<\bIFeedback\b[]> => {\\\bn\b  \btry\b {\\\bn\b    \bconst\b \bcompletion\b = \bawait\b \bclient\b.\bchat\b.\bcompletions\b.\bcreate\b({\\\bn\b      \bmessages\b: [{ \brole\b: '\buser\b', \bcontent\b: \bprompt\b }],\\\bn\b      \bmodel\b,\\\bn\b      \btemperature\b,\\\bn\b    });\\\bn\b    \bconst\b \bcontent\b = \bcompletion\b.\bchoices\b[\b0\b].\bmessage\b.\bcontent\b;\\\bn\b    \breturn\b \bparseJson\b(\bcontent\b \bas\b \bstring\b);\\\bn\b  } \bcatch\b (\berror\b) {\\\bn\b    \blogger\b.\berror\b('\bFailed\b \bto\b \bparse\b \bJSON\b \bresponse\b', \berror\b);\\\bn\b    \bthrow\b \bnew\b \bError\b('\bFailed\b \bto\b \bparse\b \bJSON\b \bresponse\b');\\\bn\b  }\\\bn\b},\\\bn\b\\`\\`\\`\"\\bn\b}\\bn\b\\`\\`\\`
ERROR   Error in processing prompt

 SyntaxError  JSON Parse error: Unrecognized token '\', 7, 39
error stack:
  • 

  • 

  • native      processTicksAndRejections
        native:7
DEBUG   Markdown report:



and I don't know how to use test:llm.

Hope to get your help and advice.