/setup

Hindsight Setup Skill

When to Use This Skill

Run this once on first use of the plugin (or any time you want to change the LLM provider Hindsight uses for its memory operations). It walks you through choosing a provider and model, then writes a config file that the container-startup script reads when the Hindsight server is next created.

How to Execute

Perform the following steps IN ORDER. This is an interactive wizard, so ask each question and wait for the answer before moving on.

Steps

1. Check Docker availability

Run a quick check that Docker exists and the daemon is reachable:

command -v docker && docker info >/dev/null 2>&1 && echo "docker: ok" || echo "docker: unavailable"

If Docker is missing or the daemon is down, clearly WARN the user that the plugin needs Docker to run the Hindsight server, and that they should install or start Docker before they expect memory features to work. Do NOT hard-fail: continue the wizard so the config is still written and ready for later.

2. Ask for the LLM provider

Use AskUserQuestion to ask which provider Hindsight should use. Offer these options:

• OpenAI (cloud)
• Anthropic (cloud)
• Gemini (cloud)
• Groq (cloud)
• Ollama (LOCAL)
• LM Studio (LOCAL)

Note that Ollama and LM Studio run a model locally on the user's machine; the other four are cloud providers that need an API key.

3. Ask for the model

Ask which model to use, presenting a sensible default for the chosen provider. Make clear these are suggestions the user can override with any model their provider supports:

• OpenAI → gpt-5-nano
• Anthropic → claude-haiku-4-5
• Gemini → gemini-2.0-flash
• Groq → llama-3.3-70b-versatile
• Ollama → llama3.1
• LM Studio → no fixed default; ask which model the user currently has loaded

4. Ask for the base URL — LOCAL providers only

Only for Ollama and LM Studio, ask for the LLM base URL with these suggested defaults:

• Ollama → http://localhost:11434/v1
• LM Studio → http://localhost:1234/v1

Skip this question entirely for cloud providers (OpenAI, Anthropic, Gemini, Groq).

5. Ask for the API key — CLOUD providers only

Only for OpenAI, Anthropic, Gemini, and Groq, ask for the provider API key. Treat this value as sensitive: do NOT echo it back in plaintext in any summary or confirmation. Skip this question entirely for local providers (Ollama, LM Studio).

6. Write the config file

Write ~/.config/hindsight-cc/config.env. Do this carefully so a secret is never exposed in shell history or a process listing any more than necessary:

1. First create the directory with Bash:

   mkdir -p ~/.config/hindsight-cc
   

2. Write the file contents with the Write tool (NOT with echo, cat, or a heredoc that would put the API key inline on a command line). Resolve the path to an absolute path first (e.g. $HOME/.config/hindsight-cc/config.env).

   The format is plain KEY=value, one per line, no surrounding quotes. Write ONLY the keys that apply to the chosen provider:

   • Always: HINDSIGHT_API_LLM_PROVIDER and HINDSIGHT_API_LLM_MODEL
   • HINDSIGHT_API_LLM_API_KEY only for cloud providers
   • HINDSIGHT_API_LLM_BASE_URL only for local providers

   Omit any inapplicable key entirely — the parser treats an absent key as unset and falls back to the built-in default or leaves it empty.

   Example for a cloud provider (OpenAI):

   HINDSIGHT_API_LLM_PROVIDER=openai
   HINDSIGHT_API_LLM_MODEL=gpt-5-nano
   HINDSIGHT_API_LLM_API_KEY=sk-...
   

   Example for a local provider (Ollama):

   HINDSIGHT_API_LLM_PROVIDER=ollama
   HINDSIGHT_API_LLM_MODEL=llama3.1
   HINDSIGHT_API_LLM_BASE_URL=http://localhost:11434/v1
   

3. Tighten permissions and confirm the file was written, with Bash:

   chmod 600 ~/.config/hindsight-cc/config.env && ls -l ~/.config/hindsight-cc/config.env
   

   Show the ls -l output to confirm the file exists with 0600 permissions. Do NOT print the file's contents (it may contain an API key).

After

Tell the user what happens next:

• The config is read when the Hindsight container is next CREATED, not on every startup.
• If a container already exists, they may need to remove it and let the plugin recreate it: docker rm -f hindsight — or simply start a fresh Claude Code session. On the next create, ensure-hindsight.sh picks up the new config.
• The very first session before setup may have had no memory configured. That is expected; rerun this command after setup and start a new session so the container is created with the chosen provider.