mirror of
https://github.com/416rehman/DeepZero
synced 2026-08-09 11:54:39 +00:00
416rehmanandClaude Opus 5
d83d1711e0
Pipelines can now use the Claude Code you already have installed and
signed in, so LLM analysis works without a third-party API key or
per-token billing:
model: claude-code # default model
model: claude-code/sonnet # or /opus, or a full model name
DeepZero never handles your credentials - it runs your own `claude`
binary in headless mode and inherits its existing sign-in. If the CLI is
missing or not signed in, validation says so before a run starts.
The pipeline feeds untrusted decompiled code to the model, so the
integration denies tools and external servers, sends the prompt over
stdin rather than the command line, and hides any ANTHROPIC_API_KEY from
the CLI so your subscription is used and not a metered API. Usage limits
retry with backoff; sign-in failures stop immediately with a clear
message.
Under the hood, LLM backends are selected from the model string by a
registry, so another agent CLI can be added later without touching the
engine, pipelines, or stages. Existing API-key models are unaffected.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
39 lines
1.4 KiB
Bash
39 lines
1.4 KiB
Bash
# ---------------------------------------------------------
|
|
# DeepZero Environment Configuration
|
|
# ---------------------------------------------------------
|
|
# Copy this file to '.env' and populate your keys.
|
|
# DeepZero will automatically load '.env' using python-dotenv.
|
|
|
|
# =========================================================
|
|
# 1. LLM Backend
|
|
#
|
|
# Two options - pick per pipeline via `model:` in pipeline.yaml
|
|
# (or override with `deepzero run -m ...`).
|
|
#
|
|
# a) Claude Code (no API key)
|
|
# Uses your own locally installed, already-signed-in Claude
|
|
# Code CLI. Nothing to configure here - just set:
|
|
# model: claude-code # default model
|
|
# model: claude-code/sonnet # or /opus, or a full model name
|
|
# Requires: `claude` on PATH and signed in (run `claude`).
|
|
#
|
|
# b) LiteLLM (API key, metered)
|
|
# Best for CI/servers. Provide keys for whatever endpoints
|
|
# your pipelines target, e.g. model: openai/gpt-4o
|
|
# =========================================================
|
|
# GEMINI_API_KEY=AI...
|
|
# OPENAI_API_KEY=sk-...
|
|
# ANTHROPIC_API_KEY=sk-ant-...
|
|
# OPENROUTER_API_KEY=sk-or-v1-...
|
|
# VERTEXAI_PROJECT=
|
|
# VERTEXAI_LOCATION=
|
|
|
|
# =========================================================
|
|
# 2. GHIDRA Headless Configuration
|
|
# Required by the `ghidra_decompile` processor
|
|
# =========================================================
|
|
# GHIDRA_INSTALL_DIR=/path/to/ghidra_11.X_PUBLIC
|
|
# JAVA_HOME=/path/to/java/jdk-17
|
|
|
|
|