elidia-agent-cli 0.6.6

Elidia Agent CLI — Universal AI Agent for your terminal

Elidia Agent CLI

Universal AI Agent for your terminal

Code, research, create, orchestrate — with 50 LLMs, 36+ tools, and 5 execution modes.

---

What is Elidia Agent CLI?

Elidia Agent CLI is a powerful terminal-based AI agent connecting you to 50 language models through a single interface. It goes beyond chat — execute tools, run multi-model consensus, conduct deep research, generate images/video/audio, orchestrate complex workflows, and spawn parallel sub-agents (SWARM mode) — all from your terminal.

Powered by the AiUtils Developer API, Elidia Agent CLI gives you access to models from OpenAI, Anthropic, Google, DeepSeek, Meta, Alibaba, and more — with a single API key.

Key Features

🎨 Hermes-Style Terminal UI (v0.6.0)

• ASCII art logo — "ELIDIA AGENT" in ANSI Shadow at launch
• Styled prompt bar — model, mode, context usage, timing, and cost before each input
• Rich welcome panel — tools, MCP servers, execution modes, personas at a glance
• Theme support — 6 built-in themes (default, dark, light, ocean, sunset, minimal)

🤖 5 Execution Modes

• DIRECT — fast single-model answers for simple questions
• CONSENSUS — 2-3 models vote, synthesize the best response
• HARNESS — multi-step agent loop with planning, tool use, self-correction
• DEEP — extended reasoning with DeepEvaluator (8-dimension quality scoring)
• SWARM — parallel sub-agents with full tool access for complex tasks

🧠 Self-Learning System

• OutcomeTracker — records every task execution (model, success/failure, tokens, cost)
• PatternLearner — cross-session analysis for model optimization
• KnowledgeGraph — LLM-powered entity + relation extraction from conversations
• Adaptive routing — learned preferences influence model selection (30% weight)

📋 Project Orchestrator

• SCRUM-style planning — EPIC → Story → Ticket decomposition
• SprintExecutor — multi-agent parallel execution with dependency ordering
• Iterative refinement — 1-3 passes (small jobs) or 1-5 passes (large jobs)

🌐 Portal AI Tools (141+ from AiUtils.io)

• Seamless access to all 141+ AI tools from the AiUtils.io production portal
• 19 genres: image_generation, video_generation, text_generation, audio_generation, code_generation, data_analysis, pdf_tools, image_editing, voiceover, content_writing, seo_tools, email_tools, social_media, chatbot, translation, research, automation, finance, legal
• Auto-routing — just describe your task naturally, the agent finds the right tool
• Direct call — /call tool_slug --key=value to invoke any portal tool directly
• Discovery — /tools shows compact count per category, /tools <category> lists tools, /tools --full shows all

🔧 Built-in Local Tools (12)

• File system — read, write, search, manage files
• Git — status, diff, log, commit
• Terminal — shell commands with permission controls
• Web search — multi-provider (Tavily, DuckDuckGo)
• Browser — Playwright-based web automation
• Database — read-only SQL with permission controls
• Email — SMTP send + IMAP search/read
• Calendar — local .ics read/write, conflict detection
• Office — PDF, DOCX, PPTX, XLSX read/write
• File upload — PDF, DOC/DOCX, PPT/PPTX, CSV, TXT, XLS/XLSX, images, code (max 10 files, 50MB each)
• RAG — local vector search with bge-m3 1024-dim embeddings
• MCP support — connect any MCP-compatible server (5 built-in defaults included)

🔬 Research Pipeline

• 5-stage pipeline — Decompose → Search → Analyze → Synthesize → Verify
• Cross-reference verification — claims checked against search sources (not circular self-check)
• Export — Markdown or HTML reports with citations

🎨 Creative Generation

• Images — FLUX, DALL-E, Stable Diffusion (500+ models)
• Video — Kling, Minimax (500+ models)
• Speech — ElevenLabs, OpenAI TTS (100+ models)
• Music — Suno text-to-music

🧩 Memory & RAG

• 4-tier memory — SYSTEM, USER, PROJECT, SESSION
• AutoMemory — detects corrections, preferences, confirmations
• Vector search — sqlite-vec with bge-m3 1024-dim embeddings
• Auto-compaction — summarizes sessions into persistent memory
• Knowledge graph — entities and relations extracted from conversations

🆕 v0.6.0 — Adopted from Hermes Agent (222K★) + OpenCode (190K★)

• MoA Loop — Mixture of Agents: 2-3 models run in PARALLEL with independent tool access, then synthesize
• Context Compressor — auto-summarizes long conversations using cheap model, detects overflow from 25+ providers
• Tool Guardrails — pre-execution safety: blocks destructive commands, path traversal, SSRF, DB drops
• Skill Bundles — 6 built-in skills + YAML-composed bundles (/skill code-review, /bundle backend-dev)
• Provider Profiles — 6 declarative provider configs (OpenAI, Anthropic, DeepSeek, Google, Ollama, AiUtils)
• Credential Pool — multi-API-key failover per provider with usage tracking
• System Prompt Tiers — stable/context/volatile architecture, cached for performance
• Gateway Server — WebSocket + REST API for remote agent access with DT enforcement + rate limiting
• Voice Tools — speech-to-text, text-to-speech, wake word detection
• Per-Project Config — auto-loads .elidia/, AGENTS.md, CLAUDE.md from project root
• Feature Flags — 10+ env-var gated features for safe experimental rollout
• DT Billing — usage bars, balance display, purchase integration with developer console

🔐 Security & Permissions

• 4-tier system — AUTO, SESSION, EVERY_TIME, NEVER
• Pre-execution guardrails — blocks destructive commands, path traversal, SSRF attacks
• Progressive trust — auto-promotes trusted actions
• Audit logging — every tool execution logged
• Keychain storage — API keys via OS keychain, never plaintext
• Message sanitization — Unicode/surrogate cleanup before API calls

---

Installation

From PyPI (recommended)

pip install elidia-agent-cli
elidia auth login          # enter your AiUtils API key
elidia ask "Hello!"        # send your first message

If elidia command not found, add pip's bin directory to PATH:

echo 'export PATH="$HOME/.local/bin:$PATH"' >> ~/.zshrc
source ~/.zshrc

Requirements

• Python 3.11 or higher
• An AiUtils Developer API key (get one here)
• macOS 12+, Ubuntu 22.04+, or Windows 10+

---

Quick Start

# 1. Authenticate
elidia auth login

# 2. One-shot message
elidia ask "Write a Python function to reverse a string"

# 3. Interactive REPL (Hermes-style UI)
elidia

# 4. With file context
elidia ask --file report.pdf "Summarize this document"

# 5. With image analysis
elidia ask --image photo.jpg "What's in this image?"

# 6. Research mode
elidia ask --mode research "Compare PostgreSQL vs MongoDB for time-series data"

# 7. Specific model
elidia ask --model claude-sonnet-4-6 "Review this code for bugs"

Using Portal AI Tools (141+)

All 141+ AI tools from AiUtils.io are available through the CLI, Desktop, Mobile, and VS Code variants.

How to call a portal tool

Auto-routing — type your request naturally, the agent finds the right tool:

> Generate a professional headshot from my photo
> Create a logo for my startup called "NovaTech"  
> Transcribe this video and translate to Spanish
> Analyze sentiment of these customer reviews

Direct call — use /call to invoke a specific tool by its slug:

/call image_generator --prompt="sunset over mountains" --style="realistic"
/call pdf_summarizer --url="https://example.com/report.pdf"
/call voiceover_generator --text="Hello world" --voice="en-US-female"
/call logo_designer --brand_name="NovaTech" --industry="technology"

Discover available tools

/tools                     # compact summary: counts per category
/tools image_generation     # detailed list for one category  
/tools --full               # all 141+ tools with descriptions

Portal tool categories

Genre	Count	Example tools
image_generation	20+	image_generator, logo_designer, background_remover, face_swap
video_generation	10+	video_creator, video_mixer, video_translator
audio_generation	15+	text_to_speech, voice_cloner, podcast_generator, music_composer
content_writing	12+	article_writer, blog_generator, ad_copy_creator, product_description
code_generation	8+	code_generator, code_reviewer, api_designer, test_writer
data_analysis	10+	data_visualizer, sentiment_analyzer, report_generator
pdf_tools	8+	pdf_merger, pdf_converter, pdf_summarizer, pdf_editor
image_editing	8+	image_resizer, photo_enhancer, color_corrector, watermark_remover
voiceover	6+	voiceover_generator, voice_cloner, dubbing_tool
seo_tools	6+	keyword_researcher, seo_auditor, meta_tag_generator
email_tools	5+	email_template_generator, email_campaign_designer
social_media	5+	social_post_generator, hashtag_researcher, carousel_designer
translation	4+	document_translator, website_translator, subtitle_translator
chatbot	4+	chatbot_builder, intent_classifier, response_generator
research	4+	research_assistant, paper_summarizer, citation_generator
finance	4+	invoice_generator, financial_report, expense_analyzer
legal	4+	contract_reviewer, legal_document_generator, compliance_checker
automation	4+	workflow_automator, data_pipeline, api_connector
text_generation	3+	text_summarizer, paraphraser, grammar_checker

Same tools across all variants

Variant	How to access
CLI	Chat naturally or /call <tool> --key=value
Desktop	Chat naturally or /call <tool> in the chat input
Mobile	Chat naturally or /call <tool> in the chat input
VS Code	Open Elidia sidebar, chat or /call <tool>

Developer API (programmatic access)

import httpx
# List all tools by genre
r = httpx.get("https://developer.aiutils.io/v1/tools/genres",
    headers={"Authorization": "Bearer ak-dev-YOUR_KEY"})
print(r.json())  # { "genres": {...}, "tools": [...] }

# Execute a specific tool
r = httpx.post("https://developer.aiutils.io/v1/tools/image_generator/execute",
    json={"inputs": {"prompt": "a cat sitting on a cloud", "style": "digital art"}},
    headers={"Authorization": "Bearer ak-dev-YOUR_KEY"})
print(r.json())  # { "task_id": "...", "result": "..." }

---

All Commands

Command	Description
elidia	Interactive REPL with Hermes-style UI and 29 slash commands
elidia ask <msg>	One-shot message (supports --model, --mode, --file, --image)
elidia auth login/logout/status	Manage API key
elidia auth email-login/email-logout/email-status	Email credentials
elidia config show	View configuration
elidia config set <section.field> <value>	Persist preferences
elidia daemon start/stop/restart/status/init	Background daemon
elidia mcp list/health	MCP server management
elidia models	List all 50 LLMs
elidia models --local	List local Ollama models
elidia rag ingest/search/list/clear	Local RAG document store
elidia session list/new/delete	Session management
elidia version	Show version
elidia workflow run <file>	Execute YAML workflow

---

REPL Commands (29 total)

Command	Description
/model [name]	Switch model or show current
/mode [mode]	Switch mode (chat/code/research/think/create)
/think [level]	Set thinking depth (minimal/low/medium/high/max)
/persona [slug]	Switch domain persona (developer, doctor, lawyer, etc.)
/research <query>	5-stage deep research pipeline
/create image|video|speech|music <prompt>	AI media generation
/workflow <file.yaml>	Execute YAML workflow
/tools [category]	List available tools grouped by category
/mcp [disconnect]	MCP server status
/image <path> [message]	Attach image for vision analysis
/rag ingest|search|list|clear	Manage RAG document store
/memory search|save|forget|list	Persistent memory management
/history [query]	Search chat history
/sessions	List sessions
/outcomes [best|failures]	Model performance rankings, failure patterns
/analyze	Full learning report (trends, preferences, routing hints)
/knowledge [query]	Search knowledge graph for entities and relations
/balance	Check DT balance
/cost	Session cost breakdown
/budget	Budget governor status
/daemon status|start|stop	Background daemon control
/theme [name]	Switch UI theme (default/dark/light/ocean/sunset/minimal)
/cache [on|off|clear]	Response cache control
/pager [on|off]	Auto-pager for long responses
/new	Start new session (compacts current)
/clear	Clear conversation
/rules	Show project rules (.elidia/rules.md)
/trust	Show trust engine promotions
/help [command]	Show help

---

Available Models

Language Models (LLMs) — 50 total

Claude (Opus, Sonnet, Haiku), GPT (4o, 4.5, 5.5, o3, o4-mini), Gemini (2.5 Pro/Flash), DeepSeek (V4, Reasoner), Llama (4 Scout/Maverick, 3.3 70B), Qwen (3, 2.5 72B), Mistral (Large, Codestral), and more across 6 vendors (OpenAI, Anthropic, DeepSeek, PIAPI/Elidia-1, Local/Ollama, Aiutils).

Creative Models — available through /create and chat

Category	Count	Examples
Image Generation	500+	FLUX, SDXL, DALL-E, Midjourney
Video Generation	500+	Runway, Pika, Kling, Sora
Audio/Music	100+	ElevenLabs, MusicGen, Whisper
3D Generation	50+	TripoSR, Zero123++, Stable3D

Local Models — elidia models --local

Ollama models running on your machine. Zero API cost, works offline.

ollama pull llama3.2:3b
elidia models --local

---

Configuration

~/.elidia/config.toml:

[api]
base_url = "https://developer.aiutils.io/v1"
timeout_seconds = 120

[models]
default = "claude-sonnet-4-6"
code = "claude-sonnet-4-6"
reasoning = "deepseek-reasoner"
creative = "auto"

[permissions]
default_level = "session"

[budget]
session_limit_dt = 50000.0

[theme]
name = "default"

---

MCP Integration

Connect any MCP-compatible server:

// ~/.elidia/mcp.json
{
  "mcpServers": {
    "github": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-github"],
      "env": { "GITHUB_TOKEN": "ghp_..." }
    },
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/path/to/dir"]
    }
  }
}

---

Workflow Example

# research.yaml
name: competitor-analysis
steps:
  - name: research
    type: llm
    prompt: "List top 5 AI coding assistants with key features"
    output: findings

  - name: analyze
    type: llm
    prompt: "Create a comparison table:\n{{findings}}"
    output: analysis

  - name: save
    type: shell
    command: "echo '{{analysis}}' > competitor-report.md"

elidia workflow run research.yaml

---

Support

• PyPI: pypi.org/project/elidia-agent-cli
• GitHub: github.com/Elidia-Technology/elidia-agent-cli
• Email: [e-mail]

---

License

Proprietary software. Copyright 2026 Elidia Technology Pvt Ltd. All Rights Reserved. See LICENSE for full terms.