ci(framework): Add e2e agentapp test

.github/workflows/framework-agentapp-e2e.yml

@@ -0,0 +1,29 @@
+name: Framework AgentApp E2E
+
+on:
+  workflow_dispatch:
+
+env:
+  FLWR_TELEMETRY_ENABLED: 0
+
+jobs:
+  agentapp:
+    runs-on: ubuntu-22.04
+    timeout-minutes: 10
+    name: AgentApp E2E
+    defaults:
+      run:
+        working-directory: framework/e2e/e2e-agentapp
+    steps:
+      - uses: actions/checkout@v5
+      - name: Bootstrap
+        uses: ./.github/actions/bootstrap
+      - name: Install dependencies
+        run: python -m pip install --upgrade .
+      - name: Run AgentApp E2E
+        env:
+          BRAVE_API_KEY: ${{ secrets.BRAVE_API_KEY }}
+          EXA_API_KEY: ${{ secrets.EXA_API_KEY }}
+          FLWR_MODEL_API_KEY: ${{ secrets.FLWR_MODEL_API_KEY }}
+          TAVILY_API_KEY: ${{ secrets.TAVILY_API_KEY }}
+        run: ./../test_superlink.sh e2e-agentapp

framework/e2e/e2e-agentapp/README.md

@@ -0,0 +1,66 @@
+# Custom AgentApp FAB example
+
+This app demonstrates how to define a local `AgentApp`, bundle it into a FAB,
+submit it with `flwr run`, and inspect streamed run events through the Control API.
+
+## Run
+
+Start a SuperLink:
+
+```bash
+export FLWR_MODEL_API_KEY="YOUR API KEY"
+uv run --no-sync --python=3.11.14 flower-superlink \
+    --insecure \
+    --control-api-address 127.0.0.1:39093 \
+    --serverappio-api-address 127.0.0.1:39094 \
+    --database /tmp/flwr-agent-run.db \
+    --log-file /tmp/flwr-agent-superlink.log
+```
+
+Start a ServerApp SuperExec in another terminal:
+
+```bash
+export FLWR_MODEL_API_KEY="YOUR API KEY"
+export BRAVE_API_KEY="YOUR BRAVE API KEY"
+uv run --no-sync --python=3.11.14 flower-superexec \
+    --insecure \
+    --appio-api-address 127.0.0.1:39094 \
+    --plugin-type serverapp
+```
+
+For web search, set one of `BRAVE_API_KEY`, `TAVILY_API_KEY`, or `EXA_API_KEY`
+in the SuperExec terminal.
+
+Submit the local AgentApp. `flwr run` builds the FAB from this directory and
+submits it to the SuperLink:
+
+```bash
+uv run --no-sync --python=3.11.14 flwr run e2e/e2e-agentapp \
+    --run-config 'agent.input="What is the Flower federated learning framework? Answer in one sentence."'
+```
+
+Disable web search for a run with:
+
+```bash
+uv run --no-sync --python=3.11.14 flwr run e2e/e2e-agentapp \
+    --run-config 'agent.input="Say hello in one short sentence." agent.web-search=false'
+```
+
+You can also write the FAB file explicitly:
+
+```bash
+uv run --no-sync --python=3.11.14 flwr build --app e2e/e2e-agentapp
+```
+
+After `flwr run` prints the run ID, stream task events:
+
+```bash
+grpcurl -plaintext \
+    -import-path proto \
+    -proto flwr/proto/control.proto \
+    -d '{"run_id": 1}' \
+    127.0.0.1:39093 \
+    flwr.proto.Control/StreamRunEvents
+```
+
+Replace `1` with the run ID returned by `flwr run`.

framework/e2e/e2e-agentapp/e2e_agentapp/__init__.py

@@ -0,0 +1,15 @@
+# Copyright 2026 Flower Labs GmbH. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ==============================================================================
+"""Minimal custom Flower AgentApp example."""

framework/e2e/e2e-agentapp/e2e_agentapp/agent.py

@@ -0,0 +1,71 @@
+# Copyright 2026 Flower Labs GmbH. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ==============================================================================
+"""Custom AgentApp used to demonstrate local FAB bundling and execution."""
+
+
+from __future__ import annotations
+
+from typing import Any
+
+from flwr.agentapp import AgentApp, AgentSession
+from flwr.app import Context
+
+_INPUT_KEY = "agent.input"
+_INSTRUCTIONS_KEY = "agent.instructions"
+_MAX_OUTPUT_TOKENS_KEY = "agent.max-output-tokens"
+_MODEL_KEY = "agent.model"
+_WEB_SEARCH_KEY = "agent.web-search"
+
+app = AgentApp()
+
+
+@app.main()
+def main(agent: AgentSession, context: Context) -> None:
+    """Run a custom single-turn AgentApp."""
+    run_config = context.run_config
+
+    agent_input = run_config.get(_INPUT_KEY)
+    if not isinstance(agent_input, str) or not agent_input.strip():
+        raise ValueError(f"`{_INPUT_KEY}` must be a non-empty string.")
+
+    model = run_config.get(_MODEL_KEY)
+    if not isinstance(model, str) or not model.strip():
+        raise ValueError(f"`{_MODEL_KEY}` must be a non-empty string.")
+
+    instructions = run_config.get(_INSTRUCTIONS_KEY)
+    if not isinstance(instructions, str):
+        raise ValueError(f"`{_INSTRUCTIONS_KEY}` must be a string.")
+
+    max_output_tokens = run_config.get(_MAX_OUTPUT_TOKENS_KEY)
+    if not isinstance(max_output_tokens, int) or max_output_tokens <= 0:
+        raise ValueError(f"`{_MAX_OUTPUT_TOKENS_KEY}` must be a positive integer.")
+
+    use_web_search = run_config.get(_WEB_SEARCH_KEY)
+    if not isinstance(use_web_search, bool):
+        raise ValueError(f"`{_WEB_SEARCH_KEY}` must be a boolean.")
+
+    request: dict[str, Any] = {
+        "model": model,
+        "input": agent_input,
+        "stream": True,
+        "max_output_tokens": max_output_tokens,
+    }
+    if instructions:
+        request["instructions"] = instructions
+    if use_web_search:
+        request["tools"] = ["web_search"]
+        request["tool_choice"] = "required"
+
+    agent.responses.create(request)

framework/e2e/e2e-agentapp/pyproject.toml

@@ -0,0 +1,30 @@
+[build-system]
+requires = ["hatchling"]
+build-backend = "hatchling.build"
+
+[project]
+name = "e2e-agentapp"
+version = "0.1.0"
+description = "Minimal custom Flower AgentApp example"
+license = "Apache-2.0"
+dependencies = ["flwr @ {root:parent:parent:uri}"]
+
+[tool.hatch.build.targets.wheel]
+packages = ["e2e_agentapp"]
+
+[tool.hatch.metadata]
+allow-direct-references = true
+
+[tool.flwr.app]
+publisher = "flwrlabs"
+fab-include = ["e2e_agentapp/**/*.py"]
+
+[tool.flwr.app.components]
+agentapp = "e2e_agentapp.agent:app"
+
+[tool.flwr.app.config.agent]
+input = ""
+instructions = "Use web search before answering, then answer in one short sentence."
+model = "openai/gpt-5.5"
+max-output-tokens = 96
+web-search = true

framework/e2e/test_superlink.sh

@@ -1,6 +1,73 @@
 #!/bin/bash
 set -e
 
+if [ "$1" = "e2e-agentapp" ]; then
+  server_arg="--insecure"
+  server_app_address="127.0.0.1:9091"
+  db_arg="--database :flwr-in-memory:"
+  runtime_dependency_install_arg="--disable-runtime-dependency-installation"
+
+  # Install Flower app
+  pip install -e . --no-deps
+
+  # Remove any duplicates
+  sed -i '/^\[tool\.flwr\.federations\.e2e\]/,/^$/d' pyproject.toml
+
+  echo -e $"\n[tool.flwr.federations.e2e]\naddress = \"127.0.0.1:9093\"\ninsecure = true" >> pyproject.toml
+
+  timeout 5m flower-superlink \
+    $server_arg $db_arg $runtime_dependency_install_arg \
+    --control-api-address 127.0.0.1:9093 \
+    --serverappio-api-address "$server_app_address" &
+  sl_pid=$!
+  sleep 3
+
+  timeout 5m flower-superexec \
+    $server_arg \
+    --appio-api-address "$server_app_address" \
+    --plugin-type serverapp &
+  sx_pid=$!
+  sleep 3
+
+  # Trigger migration
+  flwr ls "." e2e || true
+
+  timeout 1m flwr run "." e2e \
+    --run-config 'agent.input="What is the Flower federated learning framework? Answer in one sentence."'
+
+  found_success=false
+  timeout=120
+  elapsed=0
+
+  cleanup_and_exit() {
+      kill $sx_pid;
+      sleep 1; kill $sl_pid;
+      exit $1
+  }
+
+  while [ "$found_success" = false ] && [ $elapsed -lt $timeout ]; do
+      output=$(flwr ls e2e --format=json)
+      status=$(echo "$output" | jq -r '.runs[0].status')
+
+      echo "Current status: $status"
+
+      if [ "$status" == "finished:completed" ]; then
+        found_success=true
+        echo "AgentApp worked correctly!"
+        cleanup_and_exit 0
+      else
+        echo "⏳ Not completed yet, retrying in 2s..."
+        sleep 2
+        elapsed=$((elapsed + 2))
+      fi
+  done
+
+  if [ "$found_success" = false ]; then
+      echo "AgentApp had an issue and timed out."
+      cleanup_and_exit 1
+  fi
+fi
+
 case "$1" in
   e2e-bare-https | e2e-bare-auth)
     ./generate.sh