docs: separate opencode vs openlama provider isolation
- opencode (global config) sees marg/fireworks only, no ocp - openlama (OPENCODE_CONFIG=openlama.json) sees ocp only via enabled_providers - Verified: opencode run -m ocp/... → Model not found; openlama run -m ocp/... → OK - openlama launcher sets OPENCODE_CONFIG + NO_PROXY + execs opencode
This commit is contained in:
Atte149
parent
c8974d9998
commit
98bbc96bf5
2 changed files with 59 additions and 32 deletions
37
AGENTS.md
37
AGENTS.md
|
|
@ -50,22 +50,35 @@ systemd/ollama-proxy.service — systemd unit
|
|||
|
||||
## Интеграция с opencode
|
||||
|
||||
В `~/.config/opencode/opencode.json` в блоке `provider`:
|
||||
**Два отдельных конфига:** `opencode` видит только marg/fireworks, `openlama` видит только ocp.
|
||||
|
||||
- `~/.config/opencode/opencode.json` — глобальный конфиг с `marg` + `fireworks` (без `ocp`)
|
||||
- `~/.config/opencode/openlama.json` — отдельный конфиг с `ocp` + `"enabled_providers": ["ocp"]`. Содержит те же instructions/mcp/plugins/agent что глобальный, но только провайдер `ocp`.
|
||||
- `~/.local/bin/openlama` — обёртка, выставляет `OPENCODE_CONFIG=~/.config/opencode/openlama.json` + `NO_PROXY=127.0.0.1,localhost,::1` + `exec opencode "$@"`
|
||||
|
||||
openlama.json (ключевая часть):
|
||||
```json
|
||||
"ocp": {
|
||||
"npm": "@ai-sdk/openai-compatible",
|
||||
"name": "Ollama Cloud (pool)",
|
||||
"options": {
|
||||
"baseURL": "http://127.0.0.1:11435/v1",
|
||||
"apiKey": "proxy"
|
||||
},
|
||||
"models": {
|
||||
"gpt-oss:20b": { "name": "gpt-oss:20b" },
|
||||
"...": { "name": "..." }
|
||||
{
|
||||
"enabled_providers": ["ocp"],
|
||||
"provider": {
|
||||
"ocp": {
|
||||
"npm": "@ai-sdk/openai-compatible",
|
||||
"name": "Ollama Cloud (pool)",
|
||||
"options": {
|
||||
"baseURL": "http://127.0.0.1:11435/v1",
|
||||
"apiKey": "proxy"
|
||||
},
|
||||
"models": {
|
||||
"gpt-oss:20b": { "name": "gpt-oss:20b" },
|
||||
"...": { "name": "..." }
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
```
|
||||
Модель выбирается через `/models` в TUI. Пул не сделан дефолтной моделью.
|
||||
`OPENCODE_CONFIG` мерджится с глобальным конфигом, но `enabled_providers: ["ocp"]` перекрывает глобальный — `marg`/`fireworks` скрыты.
|
||||
|
||||
**Модель не сделана дефолтной** — выбирается через `/models` в TUI.
|
||||
|
||||
**ВАЖНО: NO_PROXY.** Если в окружении установлен `HTTP_PROXY`/`HTTPS_PROXY` (например `http://127.0.0.1:2080`), opencode (Bun runtime) направит запросы к прокси через HTTP-прокси → 502 Bad Gateway. Нужно явно исключить localhost:
|
||||
```sh
|
||||
|
|
|
|||
54
README.md
54
README.md
|
|
@ -63,20 +63,38 @@ Configuration flags can be overridden via environment variables in the unit (or
|
|||
|
||||
## Use with OpenCode
|
||||
|
||||
Add a custom provider to `~/.config/opencode/opencode.json`:
|
||||
The pool is wired into OpenCode via a **separate config** so that `opencode` and `openlama` expose different providers:
|
||||
|
||||
| Command | Providers visible |
|
||||
|---------|-------------------|
|
||||
| `opencode` | `marg`, `fireworks` (your defaults) |
|
||||
| `openlama` | `ocp` only (Ollama Cloud pool) |
|
||||
|
||||
### How it works
|
||||
|
||||
1. The global config at `~/.config/opencode/opencode.json` contains `marg` + `fireworks` — **no `ocp`**.
|
||||
2. A dedicated config at `~/.config/opencode/openlama.json` contains `ocp` + `"enabled_providers": ["ocp"]`. `OPENCODE_CONFIG` merges with the global config, but `enabled_providers` overrides it so `marg`/`fireworks` are hidden.
|
||||
3. The `openlama` launcher (at `~/.local/bin/openlama`) sets `OPENCODE_CONFIG`, `NO_PROXY`, and execs `opencode`.
|
||||
|
||||
### openlama.json
|
||||
|
||||
```json
|
||||
"ocp": {
|
||||
"npm": "@ai-sdk/openai-compatible",
|
||||
"name": "Ollama Cloud (pool)",
|
||||
"options": {
|
||||
"baseURL": "http://127.0.0.1:11435/v1",
|
||||
"apiKey": "proxy"
|
||||
},
|
||||
"models": {
|
||||
"gpt-oss:20b": { "name": "gpt-oss:20b" },
|
||||
"gpt-oss:120b": { "name": "gpt-oss:120b" },
|
||||
"kimi-k2.6": { "name": "kimi-k2.6" }
|
||||
{
|
||||
"$schema": "https://opencode.ai/config.json",
|
||||
"enabled_providers": ["ocp"],
|
||||
"provider": {
|
||||
"ocp": {
|
||||
"npm": "@ai-sdk/openai-compatible",
|
||||
"name": "Ollama Cloud (pool)",
|
||||
"options": {
|
||||
"baseURL": "http://127.0.0.1:11435/v1",
|
||||
"apiKey": "proxy"
|
||||
},
|
||||
"models": {
|
||||
"gpt-oss:20b": { "name": "gpt-oss:20b" },
|
||||
"gpt-oss:120b": { "name": "gpt-oss:120b" }
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
```
|
||||
|
|
@ -89,22 +107,18 @@ curl -sS http://127.0.0.1:11435/v1/models | jq -r '.data[].id'
|
|||
|
||||
### NO_PROXY (important)
|
||||
|
||||
If your environment has `HTTP_PROXY`/`HTTPS_PROXY` set (e.g. `http://127.0.0.1:2080`), opencode's Bun runtime will route requests to the proxy *through* the HTTP proxy, resulting in `502 Bad Gateway`. Exclude localhost from proxying:
|
||||
If your environment has `HTTP_PROXY`/`HTTPS_PROXY` set (e.g. `http://127.0.0.1:2080`), opencode's Bun runtime will route requests to the proxy *through* the HTTP proxy, resulting in `502 Bad Gateway`. The `openlama` launcher sets this for you; for manual launches:
|
||||
|
||||
```sh
|
||||
# ~/.config/environment.d/no-proxy-local.conf
|
||||
NO_PROXY=127.0.0.1,localhost,::1
|
||||
no_proxy=127.0.0.1,localhost,::1
|
||||
NO_PROXY=127.0.0.1,localhost,::1 OPENCODE_CONFIG=~/.config/opencode/openlama.json opencode
|
||||
```
|
||||
|
||||
Or launch opencode with `NO_PROXY=127.0.0.1,localhost opencode`.
|
||||
|
||||
### `openlama` launcher
|
||||
|
||||
A convenience wrapper is installed at `~/.local/bin/openlama`. It sets `NO_PROXY`/`no_proxy` for localhost, tries to start `ollama-proxy` if it's down, and execs `opencode` with all forwarded args:
|
||||
A convenience wrapper is installed at `~/.local/bin/openlama`. It sets `NO_PROXY`/`no_proxy` for localhost, sets `OPENCODE_CONFIG` to the ocp-only config, tries to start `ollama-proxy` if it's down, and execs `opencode` with all forwarded args:
|
||||
|
||||
```sh
|
||||
openlama # TUI with ocp available as a provider
|
||||
openlama # TUI with only ocp available
|
||||
openlama run -m ocp/gpt-oss:120b "hello" # non-interactive, specific model
|
||||
```
|
||||
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue