Chips

The chip system is what makes Synthezer unique. Chips are behavioral instructions that configure how the AI thinks, acts, and delivers results — before it ever starts working on your request.

Think of chips like configuring an AI's personality, rules, and toolset. Instead of hoping the AI understands what you want, you explicitly tell it how to behave. A chip named PRECISION doesn't just label the AI as "precise" — it sends a direct instruction: "Prioritize correctness over creativity. Verify claims against provided context and flag uncertainties."

What Are Chips?

Every chip has two parts:

• Name — A short, uppercase label you see in the UI (e.g., PRECISION, NO SPECULATION).
• Description — The actual instruction the AI receives. This is what matters. The description IS the behavioral command that shapes the AI's output.

When you select a chip, its description is injected directly into the AI prompt. The AI reads it as a rule it must follow. For example:

This means chip descriptions are not metadata or labels — they are live instructions. If you edit a chip's description, you directly change how the AI behaves when that chip is active.

Synthezer ships with 180+ built-in chips covering software engineering, writing, research, data analysis, customer support, education, product management, and more.

The 5 Categories

Chips are organized into five categories. Each category serves a distinct role in shaping the AI's behavior. When sent to the AI, each category gets its own labeled section in the prompt so the AI understands the context of each instruction.

How Chips Work in the Pipeline

You select chips once in Stage 1 (Input), but they travel through the entire pipeline. Chips are included in every AI prompt from Stage 2 through Stage 5.

Two Prompt Formats

Synthezer uses two formats for sending chips to the AI, depending on the stage:

**AI MIND (Behavioral approach):**
• PRECISION: Prioritize correctness over creativity. Verify claims against provided context and flag uncertainties.
• MINIMAL DIFF: Make the smallest safe change to meet requirements. Avoid unrelated refactors.

**NO-GO RULES (Never do these):**
• NO BREAKING CHANGES: Do not introduce breaking changes without explicit approval.
• NO UNSAFE IO: Do not run destructive commands or make unsafe file changes.

**AI Mind:** PRECISION: Prioritize correctness over creativity..., MINIMAL DIFF: Make the smallest safe change...
**No-Go Rules:** NO BREAKING CHANGES: Do not introduce breaking changes..., NO UNSAFE IO: Do not run destructive commands...

Chips Across Pipeline Stages

Priority Chips

You can mark a chip as high priority. When a chip is high priority, it appears in the prompt with a (PRIMARY) tag, signaling to the AI that this instruction takes precedence over others.

Using Chips

Selecting Chips in Stage 1

When you create a new pipeline, Stage 1 shows five chip containers — one for each category. To add chips:

1. Click on a category container (e.g., "AI Mind" or "Tools").
2. Browse the chip library for that category. Chips are sorted by usage count, so your most-used chips appear first.
3. Click a chip to add it to the container. Click again to remove it.
4. Repeat for each category as needed.

You don't need to add chips to every category. Only select what's relevant to your task. An empty category simply sends "None specified" to the AI.

Chip Packs

Chip packs are pre-configured sets of chips across all five categories. Instead of selecting chips one by one, you can load a pack to populate all categories at once. Synthezer includes five default packs:

You can create your own custom packs and save them for reuse. Packs can also be renamed, updated, or deleted (default packs can be reset to their original state).

Creating Custom Chips

The built-in library covers common use cases, but the real power of chips comes from creating your own. Custom chips let you encode your specific workflows, standards, and rules.

How to Create a Chip

1. Open the chip library for any category.
2. Click the "Add Chip" button.
3. Enter a name (automatically uppercased, must be unique within the category).
4. Enter a description — this is the instruction the AI will follow.
5. Save the chip. It's now available in your library.

Naming Rules

• Names are automatically converted to uppercase.
• Names must be unique within their category. You can have a chip named "SECURITY" in both AI Mind and Implementation, but not two in the same category.
• Names use spaces, not underscores. Write "SECURITY AWARE", not "SECURITY_AWARE".
• Keep names short and descriptive — they serve as quick labels in the UI.
• Maximum length: 100 characters.

Writing Good Descriptions

The description is the most important part of a chip. It's the actual instruction the AI reads and follows. Tips:

• Be direct and imperative. Write "Do X" or "Never Y", not "The AI should consider X".
• Be specific. "Check for SQL injection in all user inputs" is better than "Be secure".
• Include what NOT to do. "Use TypeScript. Do not use any type" is clearer than just "Use TypeScript".
• Keep it under 200 characters. The AI processes many chips at once — shorter descriptions are more reliably followed.
• Maximum length: 1000 characters per description.

USE REACT FUNCTIONAL COMPONENTS WITH HOOKS. DO NOT USE CLASS COMPONENTS. ALL COMPONENTS MUST HAVE TYPESCRIPT PROPS INTERFACES.

Use React

Editing Chip Descriptions

You can modify the description of any chip — including built-in ones — to fine-tune the AI's behavior.

1. Open the chip library and find the chip you want to edit.
2. Click the chip to expand its details.
3. Edit the description text.
4. Save your changes.

Changes take effect immediately for all new pipeline runs. Existing pipelines that already used the chip will continue with whatever description was active at the time.

If you want to revert all built-in chips to their original descriptions, use the Reset Chips to Default option in the Profile tab. This will re-seed the original 180+ chip definitions without deleting your custom chips.

Importing & Exporting

Bulk Import

You can import chips in bulk using a JSON structure organized by category:

{
  "ai_mind": [
    { "name": "MY CUSTOM MIND", "description": "Think step by step. Show your reasoning." }
  ],
  "prerequisite": [
    { "name": "VERIFY API DOCS", "description": "Read the API documentation before making calls." }
  ],
  "implementation": [],
  "tools": [],
  "no_go": [
    { "name": "NO RAW SQL", "description": "Never write raw SQL. Use the ORM exclusively." }
  ]
}

Import behavior: if a chip with the same name already exists in the same category, it is left unchanged (INSERT OR IGNORE). This means importing is safe and won't overwrite your customizations.

Chip Packs

Chip packs bundle a selection of chips across all five categories into a named preset. Packs reference chips by name — they don't duplicate chip definitions.

{
  "name": "My Custom Pack",
  "chips": {
    "ai_mind": ["PRECISION", "STRUCTURE DRIVEN"],
    "prerequisite": ["READ CODEBASE", "CHECK DEPENDENCIES"],
    "implementation": ["CODE GEN", "TESTS"],
    "tools": ["TERMINAL", "FILE SYSTEM", "DIFF TOOL"],
    "no_go": ["NO BREAKING CHANGES", "NO UNSAFE IO"]
  }
}

Pack names must be unique. Default packs are protected from deletion but can be renamed or have their chip selections updated. You can reset all default packs to their original state at any time.

Profile Export

Your entire chip library and chip packs are included in profile exports. When you export your profile, all chips and packs travel with it — making it easy to back up or transfer your configuration to another machine.

Built-in Chip Library

Synthezer ships with 180+ chips across all five categories. Here is a curated selection from each:

AI Mind (40+ chips)

Prerequisites (35+ chips)

Implementation (45+ chips)

Tools (25+ chips)

No-Go (20+ chips)

Usage Tracking

Synthezer tracks how often each chip is used across your pipelines. Every time a pipeline is submitted with a chip selected, that chip's usage count increments.

• Chips are sorted by usage count in the library, so your most-used chips float to the top.
• Usage counts help you identify which chips are part of your core workflow.
• You can reset all usage counts to zero from the command bar or Profile tab if you want a fresh start.

Tips & Best Practices

API Reference

For advanced users and integrations, here are the chip-related API endpoints: