DeepSeek Prompts & Tips for DeepSeek AI Models

Write a {{language}} {{componentType}} that {{taskDescription}}.

**Technical requirements:**
- Language/runtime: {{language}} {{version}}
- Framework: {{framework}}
- Dependencies: only use {{allowedDependencies}}

**Specifications:**
- Input: {{inputSpec}}
- Output: {{outputSpec}}
- Error cases: {{errorCases}}

**Code style:**
- Follow {{styleguide}} conventions
- Add type annotations for all function signatures
- No comments unless the logic is non-obvious

Return only the code. No explanations, no markdown fences unless I ask.

Solve the following {{problemType}} problem step by step.

**Problem:**
{{problemStatement}}

**Given:**
{{givenInformation}}

**Required:**
{{requiredOutput}}

Think through this carefully before giving your final answer. For each step:
1. State what you're calculating and why
2. Show the work
3. Verify the result before moving to the next step

If there are multiple valid approaches, briefly mention alternatives but fully work through the most efficient one.

At the end, box your final answer clearly.

Complete the following task in {{targetLanguage}}.

**Task:** {{taskDescription}}

**Language requirements:**
- Write the response entirely in {{targetLanguage}}
- Use {{formalityLevel}} register (formal/informal/technical)
- Target audience: {{targetAudience}}
- If technical terms have no standard {{targetLanguage}} translation, keep the English term and provide a brief parenthetical explanation

**Context for accuracy:**
{{contextForTranslation}}

**Output format:**
{{outputFormat}}

Do not include an English translation unless I ask for one.

I'm providing a large {{contentType}} below (approximately {{tokenEstimate}} tokens). Process it according to these instructions:

**Task:** {{processingTask}}

**Processing rules:**
- Read the entire content before starting your analysis
- Reference specific sections by quoting relevant passages
- If the content contains contradictions, flag them explicitly
- Maintain accuracy — do not infer information that is not present

**Structure your response as:**
1. {{outputSection1}}
2. {{outputSection2}}
3. {{outputSection3}}
4. Confidence assessment: rate your confidence in each section (high/medium/low)

<content>
{{longContent}}
</content>

Follow these instructions precisely. Do exactly what is specified — nothing more, nothing less.

**Task:** {{taskDescription}}

**Input:**
{{input}}

**Instructions:**
1. {{step1}}
2. {{step2}}
3. {{step3}}
4. {{step4}}

**Output constraints:**
- Format: {{outputFormat}}
- Length: {{lengthConstraint}}
- Must include: {{requiredElements}}
- Must NOT include: {{excludedElements}}

**Quality check before responding:**
- Did you follow every numbered instruction?
- Does your output match the specified format exactly?
- Did you include all required elements and exclude all excluded elements?

I'm running DeepSeek {{modelVariant}} locally with the following setup:

**Infrastructure:**
- Hardware: {{hardware}}
- Quantization: {{quantization}}
- Context length: {{contextLength}}
- Inference framework: {{inferenceFramework}}

**Problem:** {{performanceProblem}}

**Current configuration:**
{{currentConfig}}

Analyze my setup and recommend optimizations:
1. **Memory optimization** — How to reduce VRAM usage without significant quality loss
2. **Speed optimization** — Inference latency improvements
3. **Quality optimization** — Sampling parameters for my use case: {{useCase}}
4. **Configuration changes** — Specific settings to adjust with before/after expected impact

Provide exact config values I can copy-paste, not general advice.

Debug the following {{language}} code. There is a bug causing {{bugDescription}}.

**Code:**
```{{language}}
{{code}}
```

**Expected behavior:** {{expectedBehavior}}
**Actual behavior:** {{actualBehavior}}
**Error message (if any):** {{errorMessage}}

Debug this step by step:
1. Read the code and trace the execution flow
2. Identify where the actual behavior diverges from expected
3. Explain the root cause clearly
4. Provide the corrected code with a comment marking what changed and why

Do not rewrite the entire function — show only the minimal fix needed.

Extract structured data from the following {{sourceType}}:

<source>
{{sourceContent}}
</source>

**Extract into this exact JSON structure:**
```json
{{targetSchema}}
```

**Extraction rules:**
- Use null for fields not found in the source (never invent data)
- Normalize dates to ISO 8601 format ({{dateTimezone}})
- Normalize {{fieldToNormalize}} to {{normalizationRule}}
- If a field has multiple possible values, use the most specific one
- For arrays, maintain the order they appear in the source

Return only valid JSON. No explanations, no markdown formatting around the JSON.