AI Frameworks · opensquilla

opensquilla

OpenSquilla is a token-efficient Python AI agent framework that routes queries to the cheapest capable model while maintaining persistent memory, web search, and sandboxed execution. It supports multiple entry points (CLI, Web UI, chat) and integrates with 20+ LLM providers via a unified API.

Talk to DEV.co Discuss Implementation

Source: GitHub — github.com/opensquilla/opensquilla

5.4k

GitHub stars

387

Forks

Python

Primary language

Apache-2.0

License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

Field	Value
Repository	opensquilla/opensquilla
Owner	opensquilla
Primary language	Python
License	Apache-2.0 — OSI-approved
Stars	5.4k
Forks	387
Open issues	87
Latest release	v0.4.1 (2026-06-29)
Last updated	2026-07-08
Source	https://github.com/opensquilla/opensquilla

What opensquilla is

Microkernel AI agent with on-device SquillaRouter model selection, pluggable LLM provider abstraction (OpenRouter, OpenAI, Anthropic, Ollama, DeepSeek, Gemini, Qwen/DashScope et al.), persistent memory system, embedded web search, sandboxed tool dispatch, and decision logging across identical turn-loop architecture. Python 3.12+, Apache 2.0 license.

Quickstart

Get the opensquilla source

Clone the repository and explore it locally.

terminalbash

1git clone https://github.com/opensquilla/opensquilla.git2cd opensquilla3# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

Cost-Optimized Multi-Model Deployment

Route lightweight queries to smaller/cheaper models and complex reasoning to larger ones within a single budget constraint, leveraging SquillaRouter's on-device decision engine to avoid per-call overhead.

Unified Agent Interface Across Channels

Deploy identical agent behavior across CLI, Web UI, and chat integrations without duplicating tool dispatch, retry logic, or decision handling—single turn loop ensures consistency.

Tool-Augmented Local-First Workflows

Build self-improving agents with persistent memory, on-device embeddings, built-in web search, and layered sandboxing for secure tool execution without external orchestration.

Implementation considerations

Python 3.12+ required; on-device SquillaRouter adds ONNX Runtime, LightGBM, and system library dependencies (libomp on macOS, Visual C++ Runtime on Windows) that may need manual installation.
LLM provider credentials must be configured (onboarding flow); no credential validation error handling details provided—test integration early with your chosen provider(s).
Persistent memory and embeddings architecture not detailed in README; review documentation to understand data storage location, encryption, and cleanup policies before deploying to regulated environments.
Tool dispatch and sandboxing mechanism unspecified; audit sandbox escape risks and tool output filtering before exposing to untrusted users or external APIs.
Preview release status means breaking schema changes possible; pin versions tightly and monitor releases for migration guides if vendoring into production systems.

When to avoid it — and what to weigh

Proprietary LLM Lock-In Required — If your workflow demands a single vendor's model or proprietary inference stack, OpenSquilla's multi-provider abstraction may introduce unnecessary complexity.
High-Security Compliance Needs Formal Audit — Project does not declare formal security certifications or penetration testing results; use cases requiring SOC 2, HIPAA, or FedRAMP compliance require explicit review of sandbox and data-handling claims.
Production Stability at Scale (Pre-Release) — Current release is 0.5.0 Preview 2; breaking changes and API shifts are possible. Avoid if you need production SLA guarantees without active community or commercial support.
Windows Unsigned Binary Concerns — Windows builds are currently unsigned; SmartScreen and enterprise policy may block execution. Terminal install workaround available but adds friction in locked-down environments.

License & commercial use

Apache License 2.0 (Apache-2.0). Permissive OSI license permitting commercial use, modification, and distribution under standard Apache 2.0 terms (no patent grant to user modifications, indemnification clause).

Apache 2.0 permits commercial use, but no warranty or support is provided by the open-source project itself. Commercial use of the software is allowed; derivative works must retain the license. No SLA, bug-fix guarantee, or commercial support model stated in README. Organizations deploying to production should plan for self-support, fork maintenance, or negotiate separate commercial agreements if needed.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

Signal	Assessment
Maintenance	Active
Documentation	Adequate
License clarity	Clear
Deployment complexity	Moderate
DEV.co fit	Good
Assessment confidence	High

Security considerations

Sandbox design and escape mitigation not documented in README. Persistent memory and embeddings stored locally (location/encryption unspecified). Web search integration may expose queries to external services; no privacy docs provided. Tool dispatch mechanism and output filtering details absent. No formal security audit, CVE history, or responsible disclosure policy mentioned. Unsigned Windows binaries. Evaluate before processing sensitive data, PII, or running untrusted user-supplied tools.

Alternatives to consider

Anthropic Claude with tool_use

Single, highly capable model with native tool use; no multi-provider routing overhead or model-selection complexity. Trade-off: higher per-call cost, vendor lock-in, but proven production reliability.

LangChain Agent + LiteLLM

Mature agent framework with multi-provider abstraction (similar to OpenSquilla's provider layer). Trade-off: larger ecosystem, steeper learning curve, less integrated token-routing; more DIY on memory and sandboxing.

OpenAI Assistants API

Fully managed agent service with built-in memory, retrieval, and code execution sandbox. Trade-off: OpenAI-only, higher per-call costs, less customization; no local router or offline capability.

Software development agency

Build on opensquilla with DEV.co software developers

Evaluate OpenSquilla for your use case. Review the technical documentation, test multi-provider routing, and assess sandbox/memory architecture for your compliance needs. Contact Devco for architecture guidance or custom integration.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

openai-agents-python OpenLLM open-llms

Related on DEV.co

Explore the category and the services that help you build with it.

AI Frameworks AI Development API Development Custom Software Talk to DEV.co

opensquilla FAQ

Does OpenSquilla work offline?

Partially. On-device SquillaRouter and embeddings support offline routing and memory. LLM inference requires a connected provider (OpenRouter, local Ollama, etc.). Web search requires internet. Check documentation for offline-first workflows.

Can I run this in production today?

Preview release status (0.5.0rc2) means breaking changes are possible. Feasible for non-critical workloads or internal tools; not recommended for revenue-critical systems without active community/commercial support and thorough testing.

How much does the router cost?

SquillaRouter is bundled and runs locally (no additional API calls). LLM provider costs depend on chosen backend and token usage; router is designed to reduce token spend by selecting cheaper models when appropriate.

What data does OpenSquilla store?

Persistent memory and session data stored locally (default ~/.opensquilla/config.toml). On-device embeddings stored locally. Web search and LLM requests sent to configured provider. No cloud backup or telemetry mentioned; verify docs for retention and deletion policies.

Software developers & web developers for hire

Adopting opensquilla is usually one piece of a larger software development effort. As a software development agency, DEV.co provides software development services and web development expertise — pairing senior software developers and web developers with your team to design, build, and operate ai frameworks software in production.

Ready to Deploy Cost-Optimized Agents?

Talk to DEV.co Schedule a Consultation