docs/agents-llama.md teaches the legacy activation flow #848

New issue

Closed

opened 2026-04-16 10:09:52 +00:00 by dev-bot · 0 comments

dev-bot commented

2026-04-16 10:09:52 +00:00

Collaborator

Problem

docs/agents-llama.md is the canonical doc for local-model agents but teaches the legacy activation path:

Tells users to set ENABLE_LLAMA_AGENT=1 in .env.
References .env.enc (SOPS-encrypted env) as a storage option.
Describes FORGE_TOKEN_LLAMA / FORGE_PASS_LLAMA as the credential keys.

None of these reflect the current hire-an-agent + [agents.X] TOML flow. .env.enc does not exist on this deployment and sops is not installed (separate infra gap). Credential keys should be FORGE_TOKEN_<FORGE_USER_UPPER>.

Scope

Rewrite the doc around:

disinto hire-an-agent <name> <role> --local-model <url> --model <name> as the entry point.
[agents.X] TOML block as the source of truth.
Per-forge-user env var naming (FORGE_TOKEN_DEV_QWEN, etc.).
COMPOSE_PROFILES=agents-<name> requirement until #845 is resolved.
Drop .env.enc references unless/until sops is shipped as a prerequisite.

Acceptance

A new user following docs/agents-llama.md end-to-end can bring up a working llama sidecar on a fresh box.
No references to ENABLE_LLAMA_AGENT or FORGE_TOKEN_LLAMA remain.
.env.enc only mentioned if sops/.env.enc is actually supported on the factory.

Depends on: #846 (activation path must be chosen first). Related: #845, #847.

## Problem `docs/agents-llama.md` is the canonical doc for local-model agents but teaches the legacy activation path: - Tells users to set `ENABLE_LLAMA_AGENT=1` in `.env`. - References `.env.enc` (SOPS-encrypted env) as a storage option. - Describes `FORGE_TOKEN_LLAMA` / `FORGE_PASS_LLAMA` as the credential keys. None of these reflect the current `hire-an-agent` + `[agents.X]` TOML flow. `.env.enc` does not exist on this deployment and `sops` is not installed (separate infra gap). Credential keys should be `FORGE_TOKEN_<FORGE_USER_UPPER>`. ## Scope Rewrite the doc around: - `disinto hire-an-agent <name> <role> --local-model <url> --model <name>` as the entry point. - `[agents.X]` TOML block as the source of truth. - Per-forge-user env var naming (`FORGE_TOKEN_DEV_QWEN`, etc.). - `COMPOSE_PROFILES=agents-<name>` requirement until #845 is resolved. - Drop `.env.enc` references unless/until sops is shipped as a prerequisite. ## Acceptance - A new user following `docs/agents-llama.md` end-to-end can bring up a working llama sidecar on a fresh box. - No references to `ENABLE_LLAMA_AGENT` or `FORGE_TOKEN_LLAMA` remain. - `.env.enc` only mentioned if sops/`.env.enc` is actually supported on the factory. Depends on: #846 (activation path must be chosen first). Related: #845, #847.