From 7b10354962049d66bc25ea42613faea3b9eadce3 Mon Sep 17 00:00:00 2001
From: bussyjd <jd@obol.tech>
Date: Tue, 17 Mar 2026 17:09:42 +0800
Subject: [PATCH] feat: detect CLAUDE_CODE_OAUTH_TOKEN as Anthropic API key
 fallback

obolup.sh check_agent_model_api_key() now checks CLAUDE_CODE_OAUTH_TOKEN
when ANTHROPIC_API_KEY is missing, so developers with Claude Code
subscriptions skip the interactive prompt. Also documents the LLM
auto-configuration flow and OpenClaw version management in CLAUDE.md.

Closes #272 (Unit 3)
---
 CLAUDE.md | 11 +++++++++++
 obolup.sh |  7 +++++++
 2 files changed, 18 insertions(+)

diff --git a/CLAUDE.md b/CLAUDE.md
index 1de55441..e0ed19a9 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -107,6 +107,8 @@ k3d: 1 server, ports 80:80 + 8080:80 + 443:443 + 8443:443, `rancher/k3s:v1.35.1-
 
 **LiteLLM gateway** (`llm` ns, port 4000): OpenAI-compatible proxy routing to Ollama/Anthropic/OpenAI. ConfigMap `litellm-config` (YAML config.yaml with model_list), Secret `litellm-secrets` (master key + API keys). Auto-configured with Ollama models during `obol stack up` (no manual `obol model setup` needed). `ConfigureLiteLLM()` patches config + Secret + restarts. Custom endpoints: `obol model setup custom --name --endpoint --model` (validates before adding). Paid remote inference stays on vanilla LiteLLM with a static route `paid/* -> openai/* -> http://127.0.0.1:8402`; no LiteLLM fork is required. OpenClaw always routes through LiteLLM (openai provider slot), never native providers; `dangerouslyDisableDeviceAuth` is enabled for Traefik-proxied access.
 
+**Auto-configuration**: During `obol stack up`, `autoConfigureLLM()` detects host Ollama models and patches LiteLLM config so agent chat works immediately without manual `obol model setup`. During install, `obolup.sh` `check_agent_model_api_key()` reads `~/.openclaw/openclaw.json` agent model, resolves API key from environment (`ANTHROPIC_API_KEY`, `CLAUDE_CODE_OAUTH_TOKEN` for Anthropic; `OPENAI_API_KEY` for OpenAI), and exports it for downstream tools.
+
 **Per-instance overlay**: `buildLiteLLMRoutedOverlay()` reuses "ollama" provider slot pointing at `litellm.llm.svc:4000/v1` with `api: openai-completions`. App → litellm:4000 → routes by model name → actual API.
 
 ## Standalone Inference Gateway
@@ -148,6 +150,15 @@ Skills = SKILL.md + optional scripts/references, embedded in `obol` binary (`int
 4. **`OBOL_DEVELOPMENT=true`** — required for `obol stack up` to auto-build local images (x402-verifier, x402-buyer)
 5. **Root-owned PVCs** — `-f` flag required to remove in `obol stack purge`
 
+### OpenClaw Version Management
+
+Three places pin the OpenClaw version — all must agree:
+1. `internal/openclaw/OPENCLAW_VERSION` — source of truth (Renovate watches, CI reads)
+2. `internal/openclaw/openclaw.go` — `openclawImageTag` constant
+3. `obolup.sh` — `OPENCLAW_VERSION` shell constant for standalone installs
+
+`TestOpenClawVersionConsistency` in `internal/openclaw/version_test.go` catches drift.
+
 ### Pitfalls
 
 1. **Kubeconfig port drift** — k3d API port can change between restarts. Fix: `k3d kubeconfig write <name> -o .workspace/config/kubeconfig.yaml --overwrite`
diff --git a/obolup.sh b/obolup.sh
index 6b237142..28ade1e1 100755
--- a/obolup.sh
+++ b/obolup.sh
@@ -1531,6 +1531,13 @@ except: pass
 		return 0
 	fi
 
+	# Anthropic-specific fallback: Claude Code subscription token
+	if [[ "$provider" == "anthropic" && -n "${CLAUDE_CODE_OAUTH_TOKEN:-}" ]]; then
+		export ANTHROPIC_API_KEY="$CLAUDE_CODE_OAUTH_TOKEN"
+		log_success "Claude Code subscription detected (CLAUDE_CODE_OAUTH_TOKEN)"
+		return 0
+	fi
+
 	# Interactive: prompt for the API key (like hermes-agent's setup wizard)
 	if [[ -c /dev/tty ]]; then
 		log_info "Your agent uses $primary_model ($provider_name)"