Skip to main content

Introduction

The Wikis application within ELITEA is an advanced AI-powered documentation generator that analyzes code repositories (GitHub, GitLab, Bitbucket, ADO Repos) to create comprehensive wikis through intelligent code analysis and natural language processing. With Wikis, your ELITEA Agents can automatically generate documentation, answer questions about codebases, and perform deep research on repository functionality.

Automated Wiki Generation

Creates comprehensive documentation from repository analysis, including architecture overviews, API documentation, and code explanations.

Intelligent Code Analysis

Uses Enhanced Unified Graph Builder (EUGB) to understand code structure, dependencies, and relationships across the repository.

Interactive Q&A

Enables natural language questions about the codebase with context-aware answers based on the generated knowledge base.

Deep Research Capabilities

Performs multi-step analysis with planning and investigation to answer complex questions about repository functionality.

Diagram Generation

Automatically creates Mermaid diagrams to visualize architecture and technical concepts.
Integrating Wikis with ELITEA empowers your Agents to understand, document, and explain codebases automatically, making it easier to onboard new team members, maintain documentation, and answer technical questions about your projects.

Prerequisites

Code Repository Toolkit

A configured toolkit for accessing your code repository (GitHub, GitLab, Bitbucket, or ADO Repos).

LLM Configuration

Access to a configured LLM model for text generation.

Embedding Model

A configured embedding model for vector search and retrieval.

System Integration with ELITEA

Wikis integrates with ELITEA as an Application accessible through the Apps menu. It provides specialized documentation generation and analysis tools with a built-in user interface for managing wiki content and repository analysis workflows.

Request Wikis Application Access

If the Wikis application is not yet enabled for your project, the App Catalog card shows a Request Access button instead of Configure. Submit a request to have it enabled by your administrator.
  1. Navigate to Apps: Open the sidebar and select Apps.
  2. Open the App Catalog tab: Select the App Catalog tab to browse available applications.
  3. Locate the Wikis card: Find the Wikis card in the catalog.
  4. Click Request Access: Click the Request Access button on the Wikis card.
  5. Provide a reason: In the request modal, enter a description of why you need access to the Wikis application.
  6. Submit the request: Click Submit to send the request to your administrator.
  7. Await approval: Once approved, the card updates and the Configure button becomes available.
  8. Proceed with configuration: Follow the steps in Create Wikis Application to set up the application. Wikis Request Access

Create Wikis Application

  1. Navigate to Apps: Open the sidebar and select Apps.
  2. Open the App Catalog tab: Select the App Catalog tab to browse available applications.
  3. Locate the Wikis card: Find the Wikis card in the catalog.
  4. Start configuration: Click the Configure button on the Wikis card. This opens the application configuration form. Wikis Application Navigation
  5. Configure Application Settings:
    ParameterRequiredDescriptionExample
    Toolkit NameYesDescriptive name for the application”Project Documentation Generator”
    DescriptionNoOptional description of the application’s purpose”Generates wiki documentation for main project repo”
    LLM ModelYesLLM model for text generation and wiki content creation”gpt-4”, “claude-sonnet-3-5”
    Embedding ModelNoEmbedding model for vector search and retrieval (default: text-embedding-3-small)“text-embedding-3-small”
    Code ToolkitYesCode toolkit (GitHub, GitLab, Bitbucket, or ADO Repos) for accessing repository dataSelect from configured toolkits
    Max TokensYesMaximum tokens for LLM responses (default: 64000)64000, 128000
    ToolsNoSelect which tools to enable (default: all tools)generate_wiki, ask, deep_research
  6. Save Application: Click Save to create the application. Wikis-Create-Application
The Wikis Application includes an integrated user interface accessible through the application detail page. This interface provides:
  • Visual wiki generation management
  • Progress monitoring for documentation generation
  • Access to generated wiki content
  • Repository analysis status

Available Tools

The Wikis application toolkit provides three tools:
Tool NameRequired ParametersOptional ParametersDescription
generate_wikiqueryplanner_type, exclude_testsAnalyzes the repository and generates a comprehensive structured wiki. Supports sync and async invocation.
askquestionrepo_identifier_override, analysis_key_overrideAnswers targeted questions about the repository using the generated knowledge base. Requires wiki to be generated first.
deep_researchquestionrepo_identifier_override, analysis_key_overridePerforms multi-step deep analysis with planning, delegation, and comprehensive investigation. Requires wiki to be generated first.

Applications Tab

All created Wikis applications are listed in the Applications tab of the Apps menu. Each card displays the application name and description. To open a Wikis application:
  1. Navigate to the Apps menu in the sidebar and select the Applications tab.
  2. Click on your Wikis application card.
  3. The Wikis interface opens. Here you can manage wiki generation, access generated documentation, and monitor repository analysis. Applications Tab
Use the action menu on any card to edit application settings or delete the application.
If no applications have been configured yet, the Applications tab shows an empty state with the message “No applications yet”. Click + Create to open the App Catalog, or click Start Guided Tour to take an interactive walkthrough.Applications Empty State

Working with the Wikis Application

The Wikis interface is organized into three main panels:
  • Left Sidebar: Navigation panel for browsing generated wiki pages organized by sections
  • Main Content Area: Wiki content with formatted markdown, code blocks, and diagrams
  • Chat Panel: (Right side) For asking questions about your codebase using natural language Wikis Interface

Wiki Generation Workflow

Once you’re in the Wikis application interface, follow these steps to generate documentation:
Repository analysis runs automatically in the background when you use the generate_wiki tool — no manual configuration required. Wikis uses the configured code toolkit to access repository data and processes it efficiently based on repository size. The analysis includes:
  • Code Structure Analysis: Examines repository organization, file relationships, and dependencies
  • Content Extraction: Processes source code, documentation files, and comments
  • Knowledge Graph Creation: Builds connections between different parts of your codebase
  • Vector Store Generation: Creates searchable indexes for intelligent question answering
Step 1: Initiate Wiki Generation

Check Slot Availability Before Starting

Wiki generation runs on a limited number of parallel worker slots. Before clicking Generate Wiki, check the slot availability indicator chip next to the button — if all slots are occupied, your generation will be queued until one becomes free.Slot Availability Chip
  • Green chip — slots available, generation starts immediately
  • Yellow chip — only 1 slot remaining, proceed with caution
  • Red chip — all slots full, wait for a running generation to complete before starting a new one
  1. In the left sidebar, locate and click the “Generate Wiki” button.
  2. In the “Generate Wiki Documentation?” dialog:
    • Under Structure planner, select a mode:
      • Graph clustering (recommended) — builds the outline deterministically from code-graph clusters; scales predictably from small repos to large monorepos
      • Agentic — the planner agent drafts the outline iteratively; may over-segment small or doc-light repositories
    • If Graph clustering is selected, optionally enable “Smart skip tests” to exclude test files from the cluster planner and keep the outline focused
    • Click “Update” to start generation, or “Cancel” to abort
Wiki Generation
Generation time depends on your repository size: small repos (< 100 files) take 2-5 minutes, medium repos (100-1000 files) take 5-15 minutes, and large repos (> 1000 files) may take 15-30+ minutes.
Step 2: Monitor Generation Progress Once generation starts, monitor progress through the interface:
  • Progress Bar: Appears below the Generate Wiki button
  • Status Messages: Real-time updates show current operations (e.g., “Analyzing repository structure…”, “Building knowledge graph…”)
  • Elapsed Time: Counter displays how long generation has been running Generation Progress
Generation continues in the background if you navigate away. Return to the application to check completion status.
If you need to stop an in-progress generation, click the “Stop Generation” button in the sidebar. Note that partial results may not be saved.Stop

Access Generated Documentation

Once generation completes:
  1. The left sidebar populates with navigation sections
  2. Click on any section to expand and view wiki pages
  3. Select a page to view its content in the main area
  4. Generated content includes:
    • Formatted markdown documentation
    • Code examples with syntax highlighting
    • Mermaid diagrams for visualizations
    • Cross-references between related topics
Generated Documentation

Edit Wiki Pages

Wikis allows you to edit generated documentation directly in the interface to refine or customize content: To Edit a Page:
  1. Navigate to the wiki page you want to modify
  2. Click the Edit button (pencil icon) in the top-right corner of the page
  3. The page opens in an editor with markdown formatting
  4. Make your changes to the content
  5. Click Save to apply changes or Cancel to discard
Edit Wiki Pages Editing Capabilities:
  • Markdown Editing: Full markdown support for formatting text, adding links, creating lists, and more
  • Code Blocks: Modify or add code examples with syntax highlighting
  • Diagrams: Edit Mermaid diagram definitions to update visualizations
  • Content Organization: Restructure sections, add new information, or clarify existing content
If you regenerate the wiki after making manual edits, your changes will be overwritten. Save important customizations separately or use version control to track modifications.
Use manual edits for minor corrections or additions. For significant changes, consider updating the source code comments and documentation in your repository, then regenerating the wiki to reflect those updates automatically.

Update Generated Wiki

When your repository code changes, you can regenerate the wiki to update all sections with the latest information from your codebase.
Wikis automatically detects changes in your repository and updates relevant sections. The generation process is optimized to reuse cached analysis where possible, making updates faster than initial generation.
To Update Documentation:
  1. Make changes to your repository code, comments, or documentation files
  2. Commit and push your changes to the repository
  3. Return to the Wikis application interface
  4. Click the “Generate Wiki” button in the left sidebar
  5. Confirm the regeneration in the dialog
  6. Wikis will re-analyze the repository and update all sections
Update Generated Wiki
The interface includes a wiki version selector that lists all generated versions as versioned manifests, each identified by a timestamp. Select a version to browse the wiki as it was at that point. If no manifest-based versions exist, a fallback Latest (legacy) entry is shown.
Regenerating the wiki will overwrite any manual edits you made to wiki pages. If you need to preserve custom content, save it separately before regenerating.

Wiki Artifacts Storage

When you generate a wiki, Wikis stores all generated files in the wiki-artifacts bucket under a per-wiki folder. You can access and manage these files through the Artifacts Menu in ELITEA. Bucket structure:
wiki-artifacts/
├── _registry/wikis.json              # Global registry of all wikis
└── {wiki_id}/                        # Folder per wiki (e.g. owner--repo--branch)
    ├── wiki_manifest_{version}.json  # Versioned manifest (one per generation)
    ├── wiki_pages/                   # Generated markdown pages
    │   ├── {section}/
    │   │   └── {page}.md
    │   └── README.md
    ├── analysis/
    │   └── wiki_structure.json
    ├── {hash}.faiss                  # Vector index
    ├── {hash}.docstore.bin
    ├── {hash}.bm25.sqlite
    └── combined.code_graph.gz
All generated artifacts remain available in the bucket for download, sharing, or archival purposes. Artifacts Storage

Managing Settings

The Settings panel allows you to view and modify your Wikis application configuration. Opening Settings:
  1. In the left sidebar, click the “Settings” button (below Generate Wiki button)
  2. The Settings panel opens from the right side
Settings Parameters:
ParameterEditableDescription
Max TokensYesToken limit for LLM responses (default: 64000)
Code Toolkit IDYesID of the code toolkit used to access the repository
LLM ModelNoLanguage model used for wiki generation
Embedding ModelNoEmbedding model used for vector search
The LLM Model and Embedding Model fields are read-only in the Settings panel. To change the model, delete and recreate the Wikis application with the desired model configuration.
Managing Wiki Data:
  • Delete Wiki: Removes all generated wiki artifacts from the bucket
    • Use this to clean up old documentation before regenerating
    • This action cannot be undone
Saving Changes:
  1. Modify any editable configuration field
  2. Click “Save Settings” button at the bottom
  3. Settings are immediately applied to the toolkit
Settings

Use the Chat Interface

To use the chat features, you must first generate a wiki for your repository. The chat tools query the generated documentation to answer questions.
After wiki generation completes, use the chat interface to query your repository’s documentation and get AI-powered answers about your codebase. Opening the Chat:
  1. Locate the vertical Chat tab on the right edge of the interface
  2. Click it to slide open the chat drawer
The Chat tab is disabled (greyed out) until at least one wiki generation has completed. It becomes active once wiki pages are available to query.
Chat Interface Selecting a Query Tool: Wikis provides two tools for different types of questions:
ToolBest ForResponse TimeDepth
askQuick, focused questions about specific code, APIs, or functionalityFast (seconds)Concise answers with code references
deep_researchComplex analysis requiring investigation across multiple files and conceptsSlower (minutes)Comprehensive answers with evidence and multi-step reasoning
Using the Chat:
  1. Select your preferred tool using the toggle: Ask or Research
  2. Type your question in natural language
  3. Press Enter or click the Send button
  4. The response streams in real time with thinking steps visible during processing
  5. Use the copy, regenerate, or clear controls on each response as needed
  • Be specific: Instead of “How does this work?”, ask “How does the user authentication flow work in the login module?”
  • Reference components: Mention specific files, classes, or modules you’re interested in
  • Use follow-ups: Build on previous answers with more detailed questions
  • Switch tools: Start with ask for quick answers, then use deep_research for deeper investigation
For Ask tool (Quick Queries):
  • “What is the main architecture of this application?”
  • “What APIs are available in the UserService class?”
  • “How do I configure database connections?”
  • “What is the purpose of secrets in the ELITEA configuration?”
For Research tool (Complex Analysis):
  • “How does the authentication system work across all modules?”
  • “Explain the complete data flow from API request to database”
  • “What are the security implications of the current session management?”
  • “Analyze the relationships between all microservices”

Using Wikis in Your Workflows

To give an ELITEA Agent, Chat, or Pipeline access to a generated wiki knowledge base, you need a query toolkit — a separate, lightweight read-only toolkit that connects to an existing Wikis application. The Wikis plugin provides two query toolkit types depending on your use case: Create Wikis Application (Apps menu) → Generate Wiki → Create Query Toolkit (Toolkits menu) → Add Toolkit to Agent / Chat / Pipeline
The Wikis application (created in Apps) manages repository analysis, wiki generation, and the knowledge base — it is the data store. The query toolkits (created in Toolkits) are read-only connectors that agents use to query that knowledge base. They are separate objects; you need both.
  • Wikis Query toolkit — connects to one specific Wikis application. Best when your agent always queries the same repository.
  • Wiki Query toolkit — connects to the wiki registry and can query any available wiki. Best when your agent needs to work across multiple repositories or discover which wiki to use automatically.

Step 1: Create a Query Toolkit

A Wikis application must already exist and have at least one wiki fully generated before creating a query toolkit. Complete Create Wikis Application and Wiki Generation Workflow first.
Use this toolkit when your agent always queries one specific repository. It connects directly to a named Wikis application and exposes ask and deep_research for that wiki only.Create Wikis Query Toolkit
1

Navigate to Toolkits

Open the sidebar and select Toolkits.
2

Create a New Toolkit

Click + Create Toolkit. The toolkit type selector opens.
3

Select Wikis Query

Search for or scroll to Wikis Query in the type list and click it. The configuration form opens.
4

Configure the Toolkit

Fill in the toolkit settings:
ParameterRequiredDescription
NameYesA descriptive name for this toolkit (e.g., “Backend Repo Q&A”)
DescriptionNoOptional description of what repository this connects to
Wikis ToolkitYesSelect the existing Wikis application to connect to — the dropdown shows all configured Wikis applications in the project
LLM ModelYesLLM model used to answer questions
Embedding ModelYesEmbedding model used for vector search (must match the model used during wiki generation)
ToolsYesEnable the tools the agent should use: ask, deep_research
5

Save the Toolkit

Click Save. The Wikis Query toolkit is now available to add to agents, chats, and pipelines.

Available Tools
ToolDescription
askAsk a targeted question about the repository using the generated wiki knowledge base. Returns a concise answer with code references. The Wikis application must have a wiki already generated.
deep_researchPerform multi-step deep analysis on a repository topic — uses planning, delegation, and comprehensive investigation. Slower than ask but returns thorough answers with evidence. The Wikis application must have a wiki already generated.

Step 2: Add the Toolkit to an Agent, Chat, or Pipeline

1

Navigate to the Target Menu

Open the sidebar and select Agents, Chats, or Pipelines — depending on where you want to use the query toolkit.
2

Create or Open an Item

Click + Create to create a new agent, chat, or pipeline, or click an existing item’s name to open its configuration.
3

Add the Query Toolkit

  1. Scroll to the Toolkits section of the configuration.
  2. Click the + Toolkit button.
  3. In the dropdown, select the Wikis Query or Wiki Query toolkit you just created.
  4. The toolkit is added with all its tools enabled. Add Wiki Query Toolkit to Agent
4

Write Instructions (Agents only)

In the Instructions field, tell the agent when and how to use the wiki query tools.
5

Save

Click Save. The agent, chat, or pipeline can now query the wiki knowledge base.

Best Practices

  1. Clear Query Descriptions: Provide specific, detailed queries for generate_wiki — the query shapes the structure and focus of the generated documentation
  2. Use Graph Clustering for Large Repos: Prefer the Graph clustering planner for large or complex repositories; it scales predictably and produces more consistent outlines than the Agentic planner
  3. Enable Smart Skip Tests: Enable “Smart skip tests” (Graph clustering only) when your repository has a large test suite — it keeps the wiki outline focused on production code
  4. Monitor Slots: Check the slot availability indicator before starting generation to avoid queuing delays
  1. Specific Questions: Ask focused questions rather than broad “explain everything” queries
  2. Pin to a Wiki Version: Use analysis_key_override (format: owner/repo@version_id) when querying from an agent to pin results to a specific generation and avoid inconsistency if the wiki is regenerated
  3. Progressive Refinement: Start with ask for quick answers — it runs without slot limits and responds in seconds; switch to deep_research for complex multi-file analysis
  4. Reference Generated Wiki: Browse generated wiki pages before asking questions to understand what context is available
  1. Cleanup Old Wikis: Regularly remove outdated wiki artifacts using the Delete Wiki button in Settings or the delete_wiki tool — all artifacts are stored in the shared wiki-artifacts bucket
  2. Slot Monitoring: Be aware of DEEPWIKI_MAX_PARALLEL_WORKERS limits in your environment; only generate_wiki consumes slots — ask and deep_research run without slot restrictions
  3. Share One App Across Agents: When multiple agents need to query the same repository, point them all to the same Wikis application instead of creating separate apps — the generated index is reused automatically
  4. Use repo_identifier_override for Cross-Repo Queries: When using the wikis_query toolkit, pass repo_identifier_override to direct a query at a specific repository’s wiki index rather than the default configured repo

Troubleshooting

Possible causes:
  • Code toolkit authentication invalid or expired
  • Repository or branch not found
  • Repository has no indexable source files (empty_repository error)
  • LLM API errors or insufficient token limit
  • Artifact upload failure after generation completes
Solutions:
  • Check the error message returned — it includes a category (authentication_error, repository_not_found, branch_not_found, indexing_failed, empty_repository) to indicate the specific cause
  • Verify the code toolkit is properly configured and authenticated in the Toolkits menu
  • Confirm the repository exists and the configured branch is correct
  • If the error is empty_repository, the repository contains no files the indexer can process — check that the repo has source code and the code toolkit has read access
  • Increase max_tokens in Settings if generation fails mid-way on large repositories
  • Check the wiki-artifacts bucket exists in the Artifacts menu and the service has write permissions
Possible causes:
  • Maximum parallel subprocess workers reached (DEEPWIKI_MAX_PARALLEL_WORKERS, default: 1)
  • Maximum parallel K8s Jobs reached (DEEPWIKI_MAX_CONCURRENT_JOBS, default: 3) in Jobs mode
  • Other wiki generations in progress
Solutions:
  • Wait for current generations to complete — the slot indicator turns green when a slot is free
  • Check current slot status using the slot chip next to the Generate Wiki button
  • Contact your administrator to increase DEEPWIKI_MAX_PARALLEL_WORKERS (subprocess mode) or DEEPWIKI_MAX_CONCURRENT_JOBS (K8s Jobs mode)
Possible causes:
  • Wiki has not been generated yet for the repository
  • Embedding model changed after wiki generation — causes dimension mismatch on query
  • Question outside the scope of the indexed content
Solutions:
  • Ensure generate_wiki completed successfully and wiki pages are visible in the sidebar
  • If you see an error like “Dimension mismatch for query vector. Expected 1536 dimensions but received 1024”, the embedding model was changed after the wiki was generated — delete the wiki, update the Wikis application to use the original embedding model (or regenerate with the new model), then run generate_wiki again
  • The LLM Model and Embedding Model fields are read-only in Settings — to change the embedding model, delete and recreate the Wikis application
  • Rephrase the question to match terminology used in the codebase
  • Browse generated wiki pages to understand what content was indexed
Possible causes:
  • LLM provider rate limit hit during wiki generation or querying (rate_limit error)
  • API quota exhausted for the configured model
  • Too many concurrent requests to the LLM endpoint
Solutions:
  • Wait and retry — most rate limits reset within minutes to hours
  • Check LLM provider quota usage and upgrade the plan if needed
  • Reduce max_tokens in Settings to lower per-request token consumption
  • Contact your ELITEA administrator if the LLM integration is shared across projects
Possible causes:
  • Large repository with many files being cloned and indexed
  • Network latency during repository cloning
  • LLM response times for generating wiki page content
Solutions:
  • Generation time scales with repository size — small repos take 2–5 min, large repos (1000+ files) can take 15–30+ min; monitor the progress bar and status messages
  • Filesystem-based indexing (the default) is faster than API-based — verify indexing_method is not set to github unless required
  • Enable “Smart skip tests” (Graph clustering mode) to reduce the number of files indexed
  • Use the Graph clustering planner instead of Agentic — it is more predictable and faster on large repos
Possible causes:
  • Code toolkit not properly configured or authenticated
  • Missing or expired repository access credentials
  • Invalid toolkit ID reference in the Wikis application
  • Toolkit deleted or no longer accessible
Solutions:
  • Verify the code toolkit exists and is properly configured in the Toolkits menu
  • Check that access tokens or SSH keys are valid and have at least read access to the repository
  • Ensure the toolkit has permissions to clone the target repository
  • Re-select the code toolkit in the Wikis application Settings panel
  • Test the code toolkit independently before using it with Wikis
Possible causes:
  • LLM model not accessible or API key expired
  • Embedding model changed after wiki generation — vector dimensions no longer match (Dimension mismatch error)
  • Insufficient token limit for large wiki pages
Solutions:
  • Verify the LLM model is properly configured and the API key is valid
  • Check that API key quotas have not been exceeded
  • If you changed the embedding model after generating a wiki, delete the wiki and regenerate — the stored vectors and query vectors must use the same model
  • To permanently change the embedding model, delete and recreate the Wikis application with the new model, then regenerate the wiki
  • Increase max_tokens in Settings for large repositories with complex code
Possible causes:
  • Repository too large for the available worker memory
  • Too many concurrent wiki generations consuming memory
  • Insufficient memory allocated to the DeepWiki worker pod (K8s mode)
Solutions:
  • Wait for other active generations to complete before starting a new one
  • Contact your administrator to increase worker pod memory limits (K8s deployment)
  • Use the Graph clustering planner — it uses less memory than Agentic
  • Enable “Smart skip tests” to reduce the number of files processed
Possible causes:
  • Network connectivity issues between the worker and the repository host
  • Repository authentication failures (authentication_error)
  • Repository deleted, moved, or renamed
  • Branch does not exist (branch_not_found)
  • Firewall or proxy blocking outbound Git operations
Solutions:
  • Verify the repository URL and branch name configured in the code toolkit are correct
  • Confirm the repository is accessible and has not been moved or deleted
  • Ensure the access token has at minimum read/clone permissions
  • Check for firewall or proxy rules that may block Git operations from the ELITEA worker
  • Test the code toolkit independently using a simple repository operation
Possible causes:
  • wiki-artifacts bucket does not exist or was deleted
  • Insufficient write permissions for the service to upload to the bucket
  • Storage quota exceeded
  • Network issues between the worker and the artifact storage service
Solutions:
  • Verify the wiki-artifacts bucket exists in the Artifacts menu
  • Ensure the ELITEA service account has write permissions to the wiki-artifacts bucket
  • Confirm storage quota has not been exceeded
  • Contact your administrator to verify artifact storage configuration and service account permissions

FAQ

Generation time depends on repository size and complexity:
  • Small repositories (< 100 files): 2–5 minutes
  • Medium repositories (100–1,000 files): 5–15 minutes
  • Large repositories (> 1,000 files): 15–30+ minutes
Monitor the progress bar and status messages in the sidebar during generation. Generation always runs to completion in the background — if you navigate away, return to the application to check the result.
Yes. As long as your configured code toolkit (GitHub, GitLab, Bitbucket, or ADO Repos) has valid authentication credentials and the necessary permissions to clone the repository, private repositories are supported.
Run generate_wiki again. Each generation performs a full re-analysis of the repository and creates a new versioned manifest — there is no partial or incremental update. Manual edits made to wiki pages in previous generations will be overwritten by the new generation.Previous versions remain accessible via the version selector in the wiki interface.
Generation returns a [SERVICE_BUSY] error indicating all slots are occupied. You can:
  • Wait for a running generation to complete — the slot chip next to the Generate Wiki button turns green when a slot is free
  • Contact your administrator to increase DEEPWIKI_MAX_PARALLEL_WORKERS (subprocess mode) or DEEPWIKI_MAX_CONCURRENT_JOBS (K8s Jobs mode)
Note: ask and deep_research are not affected by slot limits and continue to work while generation slots are full.
Yes. Multiple agents can use the same Wikis application for ask and deep_research operations simultaneously — these tools have no slot limits. Only generate_wiki is subject to slot limits, and concurrent calls from multiple agents may result in a [SERVICE_BUSY] response for all but the first.
Wikis works with any LLM model configured in ELITEA. For best results:
  • Use models with large context windows — the default max_tokens is 64,000, so models that support 64k+ tokens produce better wiki pages
  • GPT-4 or Claude Sonnet are recommended for complex, multi-file documentation
  • Adjust max_tokens in Settings if generation fails or produces truncated content on large repositories
Yes. The embedding model must remain consistent between wiki generation and querying. If you change the embedding model after generating a wiki, queries will fail with a dimension mismatch error. To switch models: delete the existing wiki, delete and recreate the Wikis application with the new embedding model, then regenerate the wiki.

For more information about Wikis and related features: