Codeeagle

CodeEagle — Codebase Intelligence

Use CodeEagle to understand the codebase without reading source files directly. Pick the right command based on what you need:

Decision Matrix

Node Types

File, TestFile, Package, Service, Function, TestFunction, Method, Struct, Class, Interface, Enum, Variable, Constant, Type, Module, Dependency, APIEndpoint, Document, Directory, Topic, Person, DTO, AIGuideline, DBModel, DomainModel, ViewModel

Edge Types

Structured Queries (fast, machine-friendly)

List symbols in a file

codeeagle query symbols --file <path>

Returns all functions, structs, interfaces, methods with signatures, line numbers, and export status. Use this instead of reading a source file.

Find interface definitions and implementors

codeeagle query interface --name <name>
codeeagle query interface --name "Store"

Returns the interface location, signature, and all types that implement it.

Show edges for a symbol

codeeagle query edges --node <name-or-id>
codeeagle query edges --node "RunAll" --type Calls
codeeagle query edges --node "RunAll" --type Calls --direction out

Returns all incoming/outgoing relationships for a node. Filter by --type and --direction.

Trace function call chains

codeeagle query edges --node "extractFunctionCalls" --type Calls --direction out
codeeagle query edges --node "extractFunctionCalls" --type Calls --direction in

Outgoing = what this function calls. Incoming = who calls this function.

Trace import to manifest dependency

codeeagle query edges --node "github.com/dgraph-io/badger/v4" --type DependsOn

Shows import -> manifest dep links (kind=import_to_manifest).

Find API endpoints and their callers

codeeagle query --type APIEndpoint
codeeagle query edges --node "GET /api/users" --type Consumes --direction in

Find unused code

codeeagle query unused
codeeagle query unused --type Function --language go
codeeagle query unused --include-exported --package mypackage

Finds functions/methods with no incoming Calls edges. Test functions, init(), and main() are excluded. By default excludes exported functions (they may be called externally). Use --include-exported to include them.

Show test coverage

codeeagle query coverage
codeeagle query coverage --level function
codeeagle query coverage --language go --package parser

Reports which files (default) or functions have test coverage via EdgeTests edges. Shows per-package coverage percentages.

General node search

codeeagle query --type Function --name "New*" --package embedded
codeeagle query --type Struct --language go
codeeagle query --type Dependency --name "axios"
codeeagle query --type Service
codeeagle query --type TestFile
codeeagle query --type TestFunction --language rust

Search non-code files (docs, documents, images, configs)

codeeagle query --type Document --name "*.txt"
codeeagle query --type Document --name "*.docx"
codeeagle query --type Document --name "*.pdf"
codeeagle query --type Document --name "*.xlsx"
codeeagle query --type Directory
codeeagle query --type Topic
codeeagle query edges --node "authentication" --type HasTopic
codeeagle query --type Person
codeeagle query edges --node "Dad" --type AppearsIn

Non-code files are automatically indexed as Document nodes, including:

• Office documents: DOCX, PPTX, XLSX (text extracted from ZIP/XML, pure Go)
• OpenDocument: ODT, ODS, ODP (text extracted from content.xml)
• PDF: text extracted via dslipak/pdf (pure Go)
• Text files: changelogs, CSVs, config templates, design docs
• Images: PNG, JPG, GIF, WebP, BMP, TIFF (described by LLM if available)

When a docs LLM provider is available (Ollama or Vertex AI), topics are extracted and linked via HasTopic edges. Directory hierarchy is represented as Directory nodes with Contains edges.

Semantic Search (fast, meaning-based)

Find code by meaning

codeeagle rag "<natural language query>"
codeeagle rag "authentication middleware" --limit 10
codeeagle rag "database connection" --type Function,Struct
codeeagle rag "error handling patterns" --json --package api
codeeagle rag "API endpoint routing" --edges --language go
codeeagle rag "parser" --no-docs

Finds code entities semantically related to your query using hybrid ranking: vector similarity + keyword matching + graph centrality (well-connected nodes rank higher) + code entity boost (functions, structs rank above docs). Results are deduplicated by node. Requires a vector index (codeeagle sync or codeeagle vectorindex).

Flags: --limit N, --type T, --package P, --language L, --json, --edges, --min-score S, --no-docs

Use --no-docs to exclude Document/AIGuideline nodes and focus on code entities.

AI Agents (slower, prose answers)

Impact analysis and planning

codeeagle agent plan "What would be affected if I change the Store interface?"

Use for: impact analysis, dependency mapping, scope estimation, risk assessment.

Design review

codeeagle agent design "Is this new endpoint consistent with existing API patterns?"

Use for: pattern recognition, API consistency, cross-service design checks.

General questions

codeeagle agent ask "How does the branch-aware store work?"

Use for: high-level understanding, "how does X work?" questions.

Workflow for Implementation Planning

1. codeeagle query symbols --file <path> — get exact signatures in files you'll modify
2. codeeagle query interface --name <name> — understand contracts and implementors
3. codeeagle query edges --node <name> --type Calls — map the call graph
4. codeeagle query edges --node <name> --type DependsOn — trace dependency chains
5. codeeagle query edges --node <name> --type Tests — find existing tests for a symbol
6. codeeagle query coverage --level function --package <pkg> — check test coverage gaps
7. codeeagle agent plan "<what you plan to change>" — assess impact and risk

Prerequisites

• CodeEagle must be installed and on PATH
• Project must be initialized (codeeagle init) with an indexed graph (codeeagle sync)

Tips

• Run commands concurrently. All codeeagle query and codeeagle rag commands are read-only and safe to run in parallel. When you need multiple pieces of information, launch them as parallel Bash calls in a single message rather than sequentially. For example, query symbols for 3 files simultaneously, or run rag and query edges at the same time.
• Structured queries and rag are instant; AI agents take 10-20 seconds
• Use rag for natural language search ("find code that handles X"); use query for exact lookups
• Use --type Calls --direction out to see what a function depends on
• Use --type Calls --direction in to find all callers of a function
• Use --type DependsOn to trace imports through to manifest dependencies
• Use --type Tests to find which tests cover a given file or function
• Use query unused to find dead code; query coverage for test gaps
• All query and rag commands support --json for machine-readable output
• Prefer structured queries for implementation planning, AI agents for understanding
• Supported languages: Go, Python, TypeScript, JavaScript, Java, Rust, C#, Ruby, HTML, Markdown, Makefile, Shell, Terraform, YAML
• Supported document formats: DOCX, PPTX, XLSX, ODT, ODS, ODP, PDF (plus plain text, CSV, SVG, images)

Converted and distributed by TomeVault — claim your Tome and manage your conversions.