# Context Awareness (/docs/verdent/core-features/context-awareness)

> Intelligent tracking of code, task, and conversation context



Context Awareness is Verdent's ability to keep track of code, tasks, and conversation context while understanding dependencies across your project. With up to 1M token capacity, Verdent comprehends entire medium-sized codebases.

***

## What You'll Learn [#what-youll-learn]

* Understanding context windows (200K vs 1M)
* How Verdent tracks dependencies
* Context isolation across parallel tasks
* Recognizing and resetting context overload

***

## Understanding Context Windows [#understanding-context-windows]

Verdent's context window size depends on the model being used.

<Tabs>
  <Tab title="Standard Models (200K)">
    Most models use standard `200K` context windows:

    * **Claude Sonnet 4.5** - Balanced for complex tasks
    * **Claude Haiku 4.5** - Fast and efficient
    * **GPT-5 (Beta)** - Excellent for reasoning
    * **GPT-5-Codex (Beta)** - Optimized for coding
    * **Gemini 3 Pro** - Advanced reasoning and multimodal
    * **Kimi K2 Series** - Lightning-fast for coding

    **Capacity:**

    * \~`200,000` tokens of total memory capacity
    * Sufficient for most development tasks and medium-sized projects

    **What's Included:**

    * All messages in conversation
    * File contents loaded into context
    * Tool outputs and responses
    * System prompts and instructions
    * MCP server definitions

    **Performance:**

    * Degrades significantly when approaching limits
    * Watch for signs of context overload (slower responses, less accurate outputs)
    * Reset context more frequently for optimal performance
  </Tab>

  <Tab title="Extended Context (1M)">
    Claude Sonnet 4.5 1M offers extended context (`1M` tokens) for long or complex tasks exceeding `200K` input.

    **Capacity:**

    * `1,000,000` tokens of total memory
    * 5x larger than standard models

    **Benefits:**

    * Perfect for loading entire large codebases without chunking
    * Eliminates most context management concerns for big projects
    * Work longer before hitting context limits
    * Fewer session resets needed

    **When to Use:**

    * Large codebases with 1000+ files
    * Complex multi-file refactoring across entire projects
    * Long development sessions spanning multiple related tasks
    * When you want to minimize context management overhead
  </Tab>
</Tabs>

***

## What Verdent Tracks [#what-verdent-tracks]

| Context Type           | What's Preserved                                              |
| ---------------------- | ------------------------------------------------------------- |
| **Code Context**       | File contents, structure, coding patterns, file relationships |
| **Task Context**       | Current goals, progress, decisions made, conversation history |
| **Dependency Context** | Imports, exports, function calls, configuration references    |

***

## Understanding Dependencies [#understanding-dependencies]

### How Verdent Tracks Dependencies [#how-verdent-tracks-dependencies]

Verdent understands how files in your project relate to each other:

| Relationship      | Example                                     |
| ----------------- | ------------------------------------------- |
| **Imports**       | `import { UserService } from './services'`  |
| **Exports**       | `export class UserController`               |
| **References**    | Function calls, class instantiation         |
| **Configuration** | Config files referenced by multiple modules |

***

## Context Across Parallel Tasks [#context-across-parallel-tasks]

### How Context Works in Parallel Execution [#how-context-works-in-parallel-execution]

When running multiple agents in parallel, each maintains independent context:

| Aspect                      | Behavior                                    |
| --------------------------- | ------------------------------------------- |
| **Per-Workspace Context**   | Each workspace has separate context         |
| **Per-Task Context**        | Each task has its own conversation context  |
| **No Cross-Contamination**  | Parallel agents don't share context         |
| **Project-Level Awareness** | All agents can understand project structure |

### Switching Tasks [#switching-tasks]

When you switch between tasks:

| What Happens          | Description                        |
| --------------------- | ---------------------------------- |
| **Context Preserved** | Each task's context remains intact |
| **No Confusion**      | Contexts don't mix between tasks   |

***

## Recognizing Context Overload [#recognizing-context-overload]

| Symptom                | Signs                                                                            |
| ---------------------- | -------------------------------------------------------------------------------- |
| **Response Quality**   | Less accurate responses, missing details, inconsistent outputs                   |
| **Speed Issues**       | Slower responses, longer processing delays                                       |
| **Behavioral Changes** | Forgetting earlier context, redundant questions                                  |
| **Technical**          | Sessions running for several hours, or 20+ large files (>1000 lines each) loaded |

<Warning>
  If Verdent's responses become vague or repetitive, context overload may be occurring. Reset the conversation to restore full performance.
</Warning>

***

## When to Reset Context [#when-to-reset-context]

| Scenario                 | Action                                        |
| ------------------------ | --------------------------------------------- |
| **Performance degrades** | Start fresh session when quality drops        |
| **Task transitions**     | New session for unrelated features or modules |
| **After commits**        | Commit → test → new session                   |
| **Long sessions**        | Proactively reset before context degrades     |

**Best Practice:** Complete atomic unit of work → test → commit → clear context → start fresh.

***

## Best Practices [#best-practices]

| Practice                      | Benefit                                                    |
| ----------------------------- | ---------------------------------------------------------- |
| **Use @-mentions**            | Direct Verdent to specific files (`@filename`, `@folder/`) |
| **Be specific about scope**   | Limits context to relevant areas                           |
| **Focus on specific modules** | Better performance on large codebases                      |

***

## FAQs [#faqs]

<Accordion title="Which model should I choose for my project?">
  For most projects, Claude Sonnet 4.5 (200K) is sufficient. For large codebases or long development sessions, use Claude Sonnet 4.5 1M for extended context.
</Accordion>

<Accordion title="Does 1M context cost more credits?">
  Yes. Extended context models have higher credit consumption than standard 200K models.
</Accordion>

<Accordion title="How do I reset context?">
  Start a new task to reset context. Your previous task remains accessible in the task list.
</Accordion>

<Accordion title="Will I lose my work if I reset context?">
  No. File changes are saved to your workspace. Only conversation context resets.
</Accordion>

<Accordion title="Can I see how much context I've used?">
  Yes. The Context Remaining indicator in the middle panel shows remaining capacity for the current conversation.
</Accordion>

<Accordion title="Does context persist between sessions?">
  Yes. Task history persists between sessions. You can resume tasks where you left off when you reopen Verdent.
</Accordion>

***

## See Also [#see-also]

<CardGroup cols="2">
  <Card title="Agents" icon="users" href="/docs/verdent/core-features/agents">
    Context with agents
  </Card>

  <Card title="Multitasking" icon="list-check" href="/docs/verdent/core-features/multitasking">
    Context across tasks
  </Card>

  <Card title="Context Best Practices" icon="book" href="/docs/verdent/best-practices/context">
    Effective context management
  </Card>

  <Card title="Prompts Guide" icon="message" href="/docs/verdent/best-practices/prompts">
    Writing effective prompts
  </Card>
</CardGroup>