Skip to content

Fully autonomous AI Agents system capable of performing complex penetration testing tasks

feat(docker): add new provider configurations for Qwen 3.6 35B models

Dmitry Ngec9d8eb1292a
diff --git a/Dockerfile b/Dockerfile
index 850b0449..b4ce1ede 100644
--- a/Dockerfile
+++ b/Dockerfile
@@ -175,6 +175,8 @@ COPY examples/configs/vllm-qwen3.5-27b-fp8-no-think.provider.yml /opt/pentagi/co
COPY examples/configs/vllm-qwen3.5-27b-fp8.provider.yml /opt/pentagi/conf/
COPY examples/configs/vllm-qwen3.6-27b-fp8-no-think.provider.yml /opt/pentagi/conf/
COPY examples/configs/vllm-qwen3.6-27b-fp8.provider.yml /opt/pentagi/conf/
+COPY examples/configs/vllm-qwen3.6-35b-a3b-fp8-no-think.provider.yml /opt/pentagi/conf/
+COPY examples/configs/vllm-qwen3.6-35b-a3b-fp8.provider.yml /opt/pentagi/conf/
COPY examples/configs/vllm-qwen332b-fp16.provider.yml /opt/pentagi/conf/
COPY LICENSE /opt/pentagi/LICENSE
diff --git a/examples/configs/vllm-qwen3.6-35b-a3b-fp8-no-think.provider.yml b/examples/configs/vllm-qwen3.6-35b-a3b-fp8-no-think.provider.yml
new file mode 100644
index 00000000..aaeb5175
--- /dev/null
+++ b/examples/configs/vllm-qwen3.6-35b-a3b-fp8-no-think.provider.yml
@@ -0,0 +1,193 @@
+# Qwen3.6-35B-A3B FP8 Provider Configuration - THINKING MODE (default)
+# Based on official Qwen 3.6 35B-A3B recommendations for vLLM inference
+# Architecture: 35B dense A3B-aligned model with FP8 weights
+# Context: 262K native; can be extended toward ~1M tokens via rope-scaling/YaRN in the runtime
+# Modality: text-only LLM (no vision encoder), optimized for reasoning and coding workloads
+#
+# Non-thinking mode is disabled via extra_body parameter
+# Recommended sampling parameters:
+# - General tasks: temp=0.7, top_p=0.8, top_k=20, min_p=0.0, pp=1.5, rp=1.0
+# - Reasoning tasks: temp=1.0, top_p=0.95, top_k=20, min_p=0.0, pp=1.5, rp=1.0
+
+simple:
+ model: "Qwen/Qwen3.6-35B-A3B-FP8"
+ temperature: 0.7
+ top_k: 20
+ top_p: 0.8
+ min_p: 0.0
+ presence_penalty: 1.5
+ repetition_penalty: 1.0
+ n: 1
+ max_tokens: 32768
+ extra_body:
+ chat_template_kwargs:
+ enable_thinking: false
+
+simple_json:
+ model: "Qwen/Qwen3.6-35B-A3B-FP8"
+ temperature: 0.7
+ top_k: 20
+ top_p: 0.8
+ min_p: 0.0
+ presence_penalty: 1.5
+ repetition_penalty: 1.0
+ n: 1
+ max_tokens: 32768
+ json: true
+ extra_body:
+ chat_template_kwargs:
+ enable_thinking: false
+
+primary_agent:
+ model: "Qwen/Qwen3.6-35B-A3B-FP8"
+ temperature: 1.0
+ top_k: 20
+ top_p: 0.95
+ min_p: 0.0
+ presence_penalty: 1.5
+ repetition_penalty: 1.0
+ n: 1
+ max_tokens: 32768
+ extra_body:
+ chat_template_kwargs:
+ enable_thinking: false
+
+assistant:
+ model: "Qwen/Qwen3.6-35B-A3B-FP8"
+ temperature: 1.0
+ top_k: 20
+ top_p: 0.95
+ min_p: 0.0
+ presence_penalty: 1.5
+ repetition_penalty: 1.0
+ n: 1
+ max_tokens: 32768
+ extra_body:
+ chat_template_kwargs:
+ enable_thinking: false
+
+generator:
+ model: "Qwen/Qwen3.6-35B-A3B-FP8"
+ temperature: 1.0
+ top_k: 20
+ top_p: 0.95
+ min_p: 0.0
+ presence_penalty: 1.5
+ repetition_penalty: 1.0
+ n: 1
+ max_tokens: 32768
+ extra_body:
+ chat_template_kwargs:
+ enable_thinking: false
+
+refiner:
+ model: "Qwen/Qwen3.6-35B-A3B-FP8"
+ temperature: 1.0
+ top_k: 20
+ top_p: 0.95
+ min_p: 0.0
+ presence_penalty: 1.5
+ repetition_penalty: 1.0
+ n: 1
+ max_tokens: 32768
+ extra_body:
+ chat_template_kwargs:
+ enable_thinking: false
+
+adviser:
+ model: "Qwen/Qwen3.6-35B-A3B-FP8"
+ temperature: 1.0
+ top_k: 20
+ top_p: 0.95
+ min_p: 0.0
+ presence_penalty: 1.5
+ repetition_penalty: 1.0
+ n: 1
+ max_tokens: 32768
+ extra_body:
+ chat_template_kwargs:
+ enable_thinking: false
+
+reflector:
+ model: "Qwen/Qwen3.6-35B-A3B-FP8"
+ temperature: 1.0
+ top_k: 20
+ top_p: 0.95
+ min_p: 0.0
+ presence_penalty: 1.5
+ repetition_penalty: 1.0
+ n: 1
+ max_tokens: 32768
+ extra_body:
+ chat_template_kwargs:
+ enable_thinking: false
+
+searcher:
+ model: "Qwen/Qwen3.6-35B-A3B-FP8"
+ temperature: 0.7
+ top_k: 20
+ top_p: 0.8
+ min_p: 0.0
+ presence_penalty: 1.5
+ repetition_penalty: 1.0
+ n: 1
+ max_tokens: 32768
+ extra_body:
+ chat_template_kwargs:
+ enable_thinking: false
+
+enricher:
+ model: "Qwen/Qwen3.6-35B-A3B-FP8"
+ temperature: 0.7
+ top_k: 20
+ top_p: 0.8
+ min_p: 0.0
+ presence_penalty: 1.5
+ repetition_penalty: 1.0
+ n: 1
+ max_tokens: 32768
+ extra_body:
+ chat_template_kwargs:
+ enable_thinking: false
+
+coder:
+ model: "Qwen/Qwen3.6-35B-A3B-FP8"
+ temperature: 1.0
+ top_k: 20
+ top_p: 0.95
+ min_p: 0.0
+ presence_penalty: 1.5
+ repetition_penalty: 1.0
+ n: 1
+ max_tokens: 32768
+ extra_body:
+ chat_template_kwargs:
+ enable_thinking: false
+
+installer:
+ model: "Qwen/Qwen3.6-35B-A3B-FP8"
+ temperature: 1.0
+ top_k: 20
+ top_p: 0.95
+ min_p: 0.0
+ presence_penalty: 1.5
+ repetition_penalty: 1.0
+ n: 1
+ max_tokens: 32768
+ extra_body:
+ chat_template_kwargs:
+ enable_thinking: false
+
+pentester:
+ model: "Qwen/Qwen3.6-35B-A3B-FP8"
+ temperature: 1.0
+ top_k: 20
+ top_p: 0.95
+ min_p: 0.0
+ presence_penalty: 1.5
+ repetition_penalty: 1.0
+ n: 1
+ max_tokens: 32768
+ extra_body:
+ chat_template_kwargs:
+ enable_thinking: false
diff --git a/examples/configs/vllm-qwen3.6-35b-a3b-fp8.provider.yml b/examples/configs/vllm-qwen3.6-35b-a3b-fp8.provider.yml
new file mode 100644
index 00000000..7c061fb7
--- /dev/null
+++ b/examples/configs/vllm-qwen3.6-35b-a3b-fp8.provider.yml
@@ -0,0 +1,174 @@
+# Qwen3.6-35B-A3B FP8 Provider Configuration - THINKING MODE (default)
+# Based on official Qwen 3.6 35B-A3B recommendations for vLLM inference
+# Architecture: 35B dense A3B-aligned model with FP8 weights
+# Context: 262K native; can be extended toward ~1M tokens via rope-scaling/YaRN in the runtime
+# Modality: text-only LLM (no vision encoder), optimized for reasoning and coding workloads
+#
+# Thinking mode is enabled by default (no extra_body needed)
+# Recommended sampling parameters:
+# - General tasks: temp=1.0, top_p=0.95, top_k=20, min_p=0.0, pp=1.5, rp=1.0
+# - Precise coding: temp=0.6, top_p=0.95, top_k=20, min_p=0.0, pp=0.0, rp=1.0
+#
+# Non-thinking mode is disabled via extra_body parameter
+# Recommended sampling parameters:
+# - General tasks: temp=0.7, top_p=0.8, top_k=20, min_p=0.0, pp=1.5, rp=1.0
+# - Reasoning tasks: temp=1.0, top_p=0.95, top_k=20, min_p=0.0, pp=1.5, rp=1.0
+
+simple:
+ model: "Qwen/Qwen3.6-35B-A3B-FP8"
+ temperature: 0.7
+ top_k: 20
+ top_p: 0.8
+ min_p: 0.0
+ presence_penalty: 1.5
+ repetition_penalty: 1.0
+ n: 1
+ max_tokens: 32768
+ extra_body:
+ chat_template_kwargs:
+ enable_thinking: false
+
+simple_json:
+ model: "Qwen/Qwen3.6-35B-A3B-FP8"
+ temperature: 0.7
+ top_k: 20
+ top_p: 0.8
+ min_p: 0.0
+ presence_penalty: 1.5
+ repetition_penalty: 1.0
+ n: 1
+ max_tokens: 32768
+ json: true
+ extra_body:
+ chat_template_kwargs:
+ enable_thinking: false
+
+primary_agent:
+ model: "Qwen/Qwen3.6-35B-A3B-FP8"
+ temperature: 1.0
+ top_k: 20
+ top_p: 0.95
+ min_p: 0.0
+ presence_penalty: 1.5
+ repetition_penalty: 1.0
+ n: 1
+ max_tokens: 32768
+
+assistant:
+ model: "Qwen/Qwen3.6-35B-A3B-FP8"
+ temperature: 1.0
+ top_k: 20
+ top_p: 0.95
+ min_p: 0.0
+ presence_penalty: 1.5
+ repetition_penalty: 1.0
+ n: 1
+ max_tokens: 32768
+
+generator:
+ model: "Qwen/Qwen3.6-35B-A3B-FP8"
+ temperature: 1.0
+ top_k: 20
+ top_p: 0.95
+ min_p: 0.0
+ presence_penalty: 1.5
+ repetition_penalty: 1.0
+ n: 1
+ max_tokens: 32768
+
+refiner:
+ model: "Qwen/Qwen3.6-35B-A3B-FP8"
+ temperature: 1.0
+ top_k: 20
+ top_p: 0.95
+ min_p: 0.0
+ presence_penalty: 1.5
+ repetition_penalty: 1.0
+ n: 1
+ max_tokens: 32768
+
+adviser:
+ model: "Qwen/Qwen3.6-35B-A3B-FP8"
+ temperature: 1.0
+ top_k: 20
+ top_p: 0.95
+ min_p: 0.0
+ presence_penalty: 1.5
+ repetition_penalty: 1.0
+ n: 1
+ max_tokens: 32768
+
+reflector:
+ model: "Qwen/Qwen3.6-35B-A3B-FP8"
+ temperature: 1.0
+ top_k: 20
+ top_p: 0.95
+ min_p: 0.0
+ presence_penalty: 1.5
+ repetition_penalty: 1.0
+ n: 1
+ max_tokens: 32768
+ extra_body:
+ chat_template_kwargs:
+ enable_thinking: false
+
+searcher:
+ model: "Qwen/Qwen3.6-35B-A3B-FP8"
+ temperature: 0.7
+ top_k: 20
+ top_p: 0.8
+ min_p: 0.0
+ presence_penalty: 1.5
+ repetition_penalty: 1.0
+ n: 1
+ max_tokens: 32768
+ extra_body:
+ chat_template_kwargs:
+ enable_thinking: false
+
+enricher:
+ model: "Qwen/Qwen3.6-35B-A3B-FP8"
+ temperature: 0.7
+ top_k: 20
+ top_p: 0.8
+ min_p: 0.0
+ presence_penalty: 1.5
+ repetition_penalty: 1.0
+ n: 1
+ max_tokens: 32768
+ extra_body:
+ chat_template_kwargs:
+ enable_thinking: false
+
+coder:
+ model: "Qwen/Qwen3.6-35B-A3B-FP8"
+ temperature: 0.6
+ top_k: 20
+ top_p: 0.95
+ min_p: 0.0
+ presence_penalty: 0.0
+ repetition_penalty: 1.0
+ n: 1
+ max_tokens: 32768
+
+installer:
+ model: "Qwen/Qwen3.6-35B-A3B-FP8"
+ temperature: 0.6
+ top_k: 20
+ top_p: 0.95
+ min_p: 0.0
+ presence_penalty: 0.0
+ repetition_penalty: 1.0
+ n: 1
+ max_tokens: 32768
+
+pentester:
+ model: "Qwen/Qwen3.6-35B-A3B-FP8"
+ temperature: 0.6
+ top_k: 20
+ top_p: 0.95
+ min_p: 0.0
+ presence_penalty: 0.0
+ repetition_penalty: 1.0
+ n: 1
+ max_tokens: 32768