docs(onboarding): add third v1.1 starter wave

CHANGELOG.md

@@ -11,6 +11,7 @@ The format is intentionally lightweight and human-readable. Group entries by rel
 - Added richer client usage reporting in `GET /api/stats` and the dashboard, including per-client tokens, failures, success rate, and aggregate client totals
 - Added a second wave of AI-native starter templates for Agno, Semantic Kernel, Haystack, Mastra, and Google ADK
 - Added client highlight summaries to `GET /api/stats` and the built-in dashboard for top request, token, cost, failure, and latency signals
+- Added a third wave of AI-native starter templates for AutoGen, LlamaIndex, CrewAI, PydanticAI, and CAMEL
 
 ### Changed
 

README.md

@@ -173,6 +173,7 @@ Start here for the deeper deployment details:
 - [Examples](./docs/examples)
 - [First-wave AI-native starters](./docs/AI-NATIVE-MATRIX.md#first-wave-template-set-for-v110)
 - [Second-wave AI-native starters](./docs/AI-NATIVE-MATRIX.md#second-wave-template-set)
+- [Third-wave AI-native starters](./docs/AI-NATIVE-MATRIX.md#third-wave-template-set)
 - [Security review for `v1.0.0`](./docs/SECURITY-REVIEW-v1.0.0.md)
 - [Publishing](./docs/PUBLISHING.md)
 - [Troubleshooting](./docs/TROUBLESHOOTING.md)

docs/AI-NATIVE-MATRIX.md

@@ -73,13 +73,21 @@ That set gives FoundryGate:
 
 ## Second-Wave Template Set
 
-Good candidates once the first wave is stable:
+This wave is now covered in the repo:
 
 - `Agno`
 - `Semantic Kernel`
 - `Haystack`
 - `Mastra`
 - `Google ADK`
+
+## Third-Wave Template Set
+
+The next clean bundle after the second wave is:
+
+- `AutoGen`
+- `LlamaIndex`
+- `CrewAI`
 - `PydanticAI`
 - `CAMEL`
 

docs/INTEGRATIONS.md

@@ -188,6 +188,40 @@ The second wave keeps the same integration discipline while extending FoundryGat
 - recommended header: `X-FoundryGate-Client: google-adk`
 - recommended profile name: `google-adk`
 
+## Third-wave framework starters
+
+The third wave rounds out the most visible remaining framework set from the AI-native matrix.
+
+### AutoGen
+
+- starter: [examples/autogen-foundrygate.env.example](./examples/autogen-foundrygate.env.example)
+- recommended header: `X-FoundryGate-Client: autogen`
+- recommended profile name: `autogen`
+
+### LlamaIndex
+
+- starter: [examples/llamaindex-foundrygate.env.example](./examples/llamaindex-foundrygate.env.example)
+- recommended header: `X-FoundryGate-Client: llamaindex`
+- recommended profile name: `llamaindex`
+
+### CrewAI
+
+- starter: [examples/crewai-foundrygate.env.example](./examples/crewai-foundrygate.env.example)
+- recommended header: `X-FoundryGate-Client: crewai`
+- recommended profile name: `crewai`
+
+### PydanticAI
+
+- starter: [examples/pydanticai-foundrygate.env.example](./examples/pydanticai-foundrygate.env.example)
+- recommended header: `X-FoundryGate-Client: pydanticai`
+- recommended profile name: `pydanticai`
+
+### CAMEL
+
+- starter: [examples/camel-foundrygate.env.example](./examples/camel-foundrygate.env.example)
+- recommended header: `X-FoundryGate-Client: camel`
+- recommended profile name: `camel`
+
 ## Provider onboarding
 
 When onboarding a new provider:

docs/ONBOARDING.md

@@ -160,6 +160,11 @@ Starter files:
 - [examples/haystack-foundrygate.env.example](./examples/haystack-foundrygate.env.example)
 - [examples/mastra-foundrygate.env.example](./examples/mastra-foundrygate.env.example)
 - [examples/google-adk-foundrygate.env.example](./examples/google-adk-foundrygate.env.example)
+- [examples/autogen-foundrygate.env.example](./examples/autogen-foundrygate.env.example)
+- [examples/llamaindex-foundrygate.env.example](./examples/llamaindex-foundrygate.env.example)
+- [examples/crewai-foundrygate.env.example](./examples/crewai-foundrygate.env.example)
+- [examples/pydanticai-foundrygate.env.example](./examples/pydanticai-foundrygate.env.example)
+- [examples/camel-foundrygate.env.example](./examples/camel-foundrygate.env.example)
 
 ### 3b. First-wave framework starters
 
@@ -185,6 +190,18 @@ The second post-`1.0` starter wave extends the same pattern to:
 
 Keep these on the shared OpenAI-compatible path first. The right time to split them into more specialized profiles is after traces and stats show a real difference in locality, fallback, or cost behavior.
 
+### 3d. Third-wave framework starters
+
+The third post-`1.0` starter wave closes the biggest remaining framework gaps from the matrix:
+
+- `AutoGen` -> `X-FoundryGate-Client: autogen`
+- `LlamaIndex` -> `X-FoundryGate-Client: llamaindex`
+- `CrewAI` -> `X-FoundryGate-Client: crewai`
+- `PydanticAI` -> `X-FoundryGate-Client: pydanticai`
+- `CAMEL` -> `X-FoundryGate-Client: camel`
+
+Treat these the same way as the earlier waves: stay on the shared OpenAI-compatible endpoint first, then split profiles only when traces and route previews show a real need.
+
 ### 4. Add request hooks only if needed
 
 Keep hooks opt-in and narrow. Good uses are:

docs/examples/autogen-foundrygate.env.example

@@ -0,0 +1,4 @@
+OPENAI_BASE_URL=http://127.0.0.1:8090/v1
+OPENAI_API_KEY=local
+FOUNDRYGATE_CLIENT_HEADER=autogen
+FOUNDRYGATE_PROFILE=autogen

docs/examples/camel-foundrygate.env.example

@@ -0,0 +1,4 @@
+OPENAI_BASE_URL=http://127.0.0.1:8090/v1
+OPENAI_API_KEY=local
+FOUNDRYGATE_CLIENT_HEADER=camel
+FOUNDRYGATE_PROFILE=camel

docs/examples/crewai-foundrygate.env.example

@@ -0,0 +1,4 @@
+OPENAI_BASE_URL=http://127.0.0.1:8090/v1
+OPENAI_API_KEY=local
+FOUNDRYGATE_CLIENT_HEADER=crewai
+FOUNDRYGATE_PROFILE=crewai

docs/examples/llamaindex-foundrygate.env.example

@@ -0,0 +1,4 @@
+OPENAI_BASE_URL=http://127.0.0.1:8090/v1
+OPENAI_API_KEY=local
+FOUNDRYGATE_CLIENT_HEADER=llamaindex
+FOUNDRYGATE_PROFILE=llamaindex

docs/examples/pydanticai-foundrygate.env.example

@@ -0,0 +1,4 @@
+OPENAI_BASE_URL=http://127.0.0.1:8090/v1
+OPENAI_API_KEY=local
+FOUNDRYGATE_CLIENT_HEADER=pydanticai
+FOUNDRYGATE_PROFILE=pydanticai

foundrygate/onboarding.py

@@ -503,6 +503,89 @@ def build_onboarding_report(
                 ),
             ],
         },
+        "autogen": {
+            "recommended": "autogen" in profile_names,
+            "header": "X-FoundryGate-Client: autogen",
+            "profile": "autogen",
+            "snippet": [
+                "export OPENAI_BASE_URL=http://127.0.0.1:8090/v1",
+                "export OPENAI_API_KEY=local",
+            ],
+            "notes": [
+                "Keep AutoGen on the shared OpenAI-compatible path first.",
+                (
+                    "Split assistant families into separate profiles only when traces "
+                    "show a real locality or cost split."
+                ),
+            ],
+        },
+        "llamaindex": {
+            "recommended": "llamaindex" in profile_names,
+            "header": "X-FoundryGate-Client: llamaindex",
+            "profile": "llamaindex",
+            "snippet": [
+                "export OPENAI_BASE_URL=http://127.0.0.1:8090/v1",
+                "export OPENAI_API_KEY=local",
+            ],
+            "notes": [
+                "Keep retrieval and generation on one gateway surface to start with.",
+                (
+                    "Add dedicated routing only when one index or workflow class needs "
+                    "different provider behavior."
+                ),
+            ],
+        },
+        "crewai": {
+            "recommended": "crewai" in profile_names,
+            "header": "X-FoundryGate-Client: crewai",
+            "profile": "crewai",
+            "snippet": [
+                "export OPENAI_BASE_URL=http://127.0.0.1:8090/v1",
+                "export OPENAI_API_KEY=local",
+            ],
+            "notes": [
+                "Use one stable client tag for CrewAI orchestration before splitting roles.",
+                (
+                    "Only add profile-specific routing when crew classes need distinct "
+                    "fallback, locality, or cost constraints."
+                ),
+            ],
+        },
+        "pydanticai": {
+            "recommended": "pydanticai" in profile_names,
+            "header": "X-FoundryGate-Client: pydanticai",
+            "profile": "pydanticai",
+            "snippet": [
+                "export OPENAI_BASE_URL=http://127.0.0.1:8090/v1",
+                "export OPENAI_API_KEY=local",
+            ],
+            "notes": [
+                (
+                    "Keep PydanticAI on the common OpenAI-compatible path unless a "
+                    "model API gap appears."
+                ),
+                (
+                    "Use client profiles only when tool or validation-heavy traffic "
+                    "deserves a different provider set."
+                ),
+            ],
+        },
+        "camel": {
+            "recommended": "camel" in profile_names,
+            "header": "X-FoundryGate-Client: camel",
+            "profile": "camel",
+            "snippet": [
+                "export OPENAI_BASE_URL=http://127.0.0.1:8090/v1",
+                "export OPENAI_API_KEY=local",
+            ],
+            "notes": [
+                "Start CAMEL traffic on the shared gateway path and keep one stable client tag.",
+                (
+                    "Only add narrower policies when multi-agent workloads need stronger "
+                    "provider isolation."
+                ),
+            ],
+        },
     }
 
     return {

tests/test_onboarding.py

@@ -266,6 +266,11 @@ def test_onboarding_report_marks_all_builtin_integrations_ready(tmp_path: Path):
     assert report["integrations"]["openclaw"]["recommended"] is True
     assert report["integrations"]["n8n"]["recommended"] is True
     assert report["integrations"]["cli"]["recommended"] is True
+    assert "autogen" in report["integrations"]
+    assert "llamaindex" in report["integrations"]
+    assert "crewai" in report["integrations"]
+    assert "pydanticai" in report["integrations"]
+    assert "camel" in report["integrations"]
     assert report["clients"]["matrix"][0]["name"] == "cli"
     assert report["clients"]["matrix"][0]["has_rule"] is True