Ryvos v0.2.0: 62 Tools, 14 Providers, and a Skill Registry
Ryvos v0.2.0: 62 Tools, 14 Providers, and a Skill Registry
Today we're shipping Ryvos v0.2.0 — the biggest release since launch. Here's what's new.
50 New Built-in Tools (62 total)
Ryvos v0.1.x shipped with 12 tools. v0.2.0 adds 50 more across 11 categories:
- Sessions — list, view history, send messages between sessions, spawn new ones
- Memory — get/delete memory entries, daily append-only logs
- File System — copy, move, delete, watch files, create/extract archives
- Git — status, diff, log, commit, branch, clone — no shell commands needed
- Code/Dev — format, lint, test, outline — auto-detects your language
- Network — HTTP requests, file downloads, DNS lookup, connectivity checks
- System — process list/kill, environment variables, system info, disk usage
- Data — JSON query, CSV parse, YAML/TOML convert, base64, hashing, regex, diff
- Scheduling — list/add/remove cron jobs via tools
- Database — SQLite query and schema inspection
- Communication — send notifications
Every tool is classified under the security overview. See the full Built-in Tools docs.
14 LLM Providers
We've added 12 new providers, for a total of 14:
4 dedicated implementations with native API support:
- Google Gemini
- Azure OpenAI
- Cohere v2
- AWS Bedrock (stub — full SigV4 support in v0.3.0)
10 OpenAI-compatible presets with automatic base URL and header configuration:
- Ollama, Groq, OpenRouter, Together, Fireworks, Cerebras, xAI, Mistral, Perplexity, DeepSeek
Switch providers by changing one line in your config. See the Providers guide.
Memory Flush Before Compaction
When the context window fills up, Ryvos now asks the agent to persist important information before compacting. This means long-running sessions no longer lose critical context during summarization.
Daily Append-Only Logs
The agent now maintains daily journal files at ~/.ryvos/memory/YYYY-MM-DD.md. The last 2 days are automatically injected into the agent's context, giving it continuity across sessions.
Webhooks
External services can now trigger your agent via HTTP:
curl -X POST https://your-host:18789/api/hooks/wake \
-H "Authorization: Bearer $TOKEN" \
-d '{"prompt": "Deploy failed. Investigate."}'See the Webhooks docs.
Skill Registry
Install community skills with a single command:
ryvos skill install web-scraper
ryvos skill search "database"
ryvos skill list --remoteSee the Custom Tools & Skills docs.
Non-Interactive Setup
New flags for CI/CD and Docker deployments:
ryvos init --from-env # reads RYVOS_PROVIDER, RYVOS_API_KEY, etc.
ryvos init -y --provider groq --model-id llama-3.3-70b-versatileSee CLI Reference.
Per-Agent Model Routing
Route different agents to different providers:
[agent.model_overrides.researcher]
provider = "groq"
model_id = "llama-3.3-70b-versatile"Upgrading
cargo install --path . --force
# or
curl -fsSL https://raw.githubusercontent.com/Ryvos/ryvos/main/install.sh | shYour existing config is fully backward-compatible. All new config sections are optional.
Full changelog on GitHub.