Install with an AI Agent

Install Aionis for this project.

Goal:
- Install a local Aionis Runtime in .aionis-runtime beside this project.
- Configure embeddings so stored-memory recall works.
- Start the Runtime locally.
- Verify the Runtime health endpoint.
- Print the exact base URL and next integration command for this project.
- Do not modify application source code unless I explicitly ask you to.
- Do not print API keys or write them into chat output.

Preferred defaults:
- Runtime directory: .aionis-runtime
- Runtime URL: http://127.0.0.1:3001
- Embedding provider: ask me whether to use openai, minimax, or another supported provider.
- If an API key is needed, ask me to provide it securely through the shell environment or a local .env file.
- If the port is already in use, choose the next free local port and report it.

Steps:
1. Check Node.js is 22 or newer.
2. Ask me which embedding provider to use.
3. Choose exactly one install command.
4. If I chose OpenAI embeddings and OPENAI_API_KEY is already available:
   OPENAI_API_KEY="$OPENAI_API_KEY" npx aionis setup .aionis-runtime --provider openai --yes
5. If I chose MiniMax embeddings and MINIMAX_API_KEY is already available:
   MINIMAX_API_KEY="$MINIMAX_API_KEY" npx aionis setup .aionis-runtime --provider minimax --yes
6. If I explicitly want to configure embeddings later:
   npx aionis setup .aionis-runtime --provider none --yes
7. If the required key is missing, stop and ask me for the key before running setup.
8. Start the Runtime in a persistent terminal:
   cd .aionis-runtime && npm run -s lite:start
9. Verify from another shell:
   curl http://127.0.0.1:3001/health
10. Report:
   - Runtime directory
   - Runtime base URL
   - Health status
   - Which embedding provider is active
   - Next command for SDK, HTTP, MCP, or AIFS based on my project shape

Acceptance criteria:
- .aionis-runtime exists.
- Runtime starts without build errors.
- /health returns success.
- The final answer includes the Aionis base URL.
- No API key is printed in chat.

task: install_aionis_runtime
working_directory: current_project_root
runtime_directory: .aionis-runtime
default_base_url: http://127.0.0.1:3001
node_requirement: ">=22"
installer:
  command: npx aionis setup .aionis-runtime --yes
embedding_provider:
  required_for:
    - stored_memory_recall
    - recall_backed_guide_output
  supported_common_options:
    - openai
    - minimax
  secret_handling:
    - never_print_api_keys
    - prefer_existing_shell_environment
    - otherwise_ask_user_before_writing_local_env
runtime_start:
  command: cd .aionis-runtime && npm run -s lite:start
health_check:
  command: curl http://127.0.0.1:3001/health
success_outputs:
  - runtime_directory
  - base_url
  - embedding_provider
  - health_status
  - next_integration_surface
do_not:
  - modify_application_source_without_user_approval
  - commit_secret_files
  - paste_api_keys_into_chat
  - assume_a_cloud_service_is_required

Connect this project's Agent loop to Aionis.

Use the Aionis Runtime at http://127.0.0.1:3001.

Implement the product loop:
1. Before the Agent acts, call guide with the current task query, scope, run_id,
   task_signature, agent_id, and team_id.
2. Give the Agent the compiled guide context.
3. After tool execution, call observe with the action summary, target files,
   validation result, outcome, and any handoff state.
4. After the outcome is known, call feedback using the guide_trace_id and only
   memory IDs that were exposed by that guide.
5. Call measure and snapshot for operator review.

Requirements:
- Keep scope stable for this project.
- Keep run_id stable for one execution session.
- Use distinct agent_id values for planner, worker, verifier, and reviewer if
  this is a multi-agent workflow.
- Preserve guide_trace_id for feedback attribution.
- Do not send raw secrets to Aionis.
- If the guide returns rehydrate pointers, fetch only the evidence needed for
  the current action.

Acceptance criteria:
- The Agent calls guide before acting.
- The Agent records observe after meaningful work.
- The Agent sends feedback tied to guide_trace_id.
- Aionis can produce a snapshot for the run.

curl http://127.0.0.1:3001/health

npx @aionis/mcp@latest --base-url http://127.0.0.1:3001 --scope-from workspace

npx @aionis/aifs@latest refresh \
  --base-url http://127.0.0.1:3001 \
  --scope your-project

import { createAionisClient } from "@aionis/sdk";
 
const aionis = createAionisClient({
  baseUrl: "http://127.0.0.1:3001",
  scope: "your-project",
});

Aionis install result:
- Runtime directory:
- Runtime base URL:
- Health status:
- Embedding provider:
- Integration selected:
- Files created:
- Commands to run next:
- Any unresolved issue: