Knowledge Base
Glossary & Concepts
Refusal vector
Definition of refusal vectors in LLMs and how they power abliteration.
Refusal vector ablation
How refusal vector ablation removes refusal behavior while preserving core model capability.
Residual stream
Definition of the transformer residual stream and why it matters for activation editing.
Orthogonalization
How orthogonalization removes unwanted directions in activation space.
LLM guardrails with audit logs and rollouts
AI gateway guardrails for LLMs with audit logs, shadow/canary rollouts, and policy-as-code enforcement.
Token quotas for LLM APIs (per-user, per-project)
AI gateway token quotas for LLM APIs using per-user and per-project limits with audit-ready enforcement.
Rewrite instead of refuse: improve LLM UX safely
Use AI gateway guardrails to rewrite sensitive outputs instead of refusing, with policy reasons and audit tags.
Shadow mode for AI policy changes
Run AI gateway policy changes in shadow mode to compare decisions before enforcing them.
Policy-as-code for LLM behavior
Define LLM behavior with policy-as-code rules, allowlists, denylists, and reason codes in an AI gateway.
What is an uncensored LLM?
Definition, use cases, and examples of uncensored large language models. Learn how they work and why developers use them.
What is abliteration?
Refusal vector ablation explained with diagrams and examples. Learn why abliteration is more stable than jailbreak prompts.
OpenAI Chat Completions compatible API
Definition and examples for OpenAI Chat Completions compatibility. Learn how to switch base URLs and keep the same SDKs.
OpenAI compatible API alternatives
How to evaluate OpenAI-compatible API alternatives, including pricing, retention policy, model behavior, and migration steps.
Streaming chat completions
Streaming chat completions explained with examples and guidance for OpenAI-compatible APIs.
Vision-capable LLM API
Definition and examples for sending images to vision-capable models via OpenAI-compatible chat completions.
LLM API rate limits
Definition and best practices for handling LLM API rate limits, retries, and backoff.
Zero data retention LLM API
Zero data retention LLM API explained, including no prompt/output retention and the role of operational telemetry.
Switch OpenAI Python SDK base URL
Switch your OpenAI Python SDK to abliteration.ai in one line. Change base_url and keep your existing code.
Switch OpenAI Node SDK base URL
Switch your OpenAI Node SDK to abliteration.ai in one line. Change baseURL and keep your existing TypeScript/JavaScript code.
Switch LangChain to abliteration.ai
Switch LangChain's ChatOpenAI to abliteration.ai by setting base_url. Keep your chains, agents, and tools unchanged.
Switch LlamaIndex to abliteration.ai
Switch LlamaIndex's OpenAI LLM to abliteration.ai by setting api_base. Keep your indexes, retrievers, and query engines unchanged.
Switch Vercel AI SDK to abliteration.ai
Switch Vercel AI SDK to abliteration.ai using the OpenAI-compatible provider. Keep your generateText and streamText calls unchanged.
Refusal replacement playbook
Replace 'I can't help with that' with policy-driven rewrites, redactions, escalations, and structured refusals. Complete playbook with reason codes and examples.
Policy template: PII redaction + safe rewrite
Ready-to-use policy template for PII redaction combined with safe rewrites. Copy, customize, and deploy.
Policy template: Customer support rewrite mode
Ready-to-use policy template for customer support that rewrites unhelpful responses into actionable guidance.
Policy template: Summarize without actionable detail
Ready-to-use policy template for high-risk topics that provides summaries without actionable instructions.