feat(agents): z.ai GLM coding plan provider + paperclip env passthrough
- opencode.nix: add zai-coding-plan provider (OpenAI-compatible coding-plan
endpoint, apiKey via {env:ZAI_API_KEY}) with glm-5.3 (1M ctx) and
glm-5-turbo — resolves in OpenCode itself, so Paperclip's opencode-harness
agents can run on the GLM flat rate
- paperclip.nix: optional EnvironmentFile=%h/.config/paperclip.env so
agents spawned by the service inherit provider keys
This commit is contained in:
1 parent
800a356b66
commit
a660e920b6
2 files changed
+36
No files matched your search
@@ -168,6 +168,38 @@
|
||||
|
||||
# AZ-Gruppe LiteLLM endpoint + available models
|
||||
provider = {
|
||||
# Z.AI GLM Coding Plan (flat-rate subscription).
|
||||
# OpenAI-compatible endpoint; API key from the z.ai console is
|
||||
# injected via {env:...} so it never lands in the store. Consumed by
|
||||
# interactive opencode2 sessions (shell env) and by agents spawned
|
||||
# through Paperclip's opencode harness (service EnvironmentFile).
|
||||
zai-coding-plan = {
|
||||
npm = "@ai-sdk/openai-compatible";
|
||||
name = "Z.AI Coding Plan";
|
||||
options = {
|
||||
baseURL = "https://api.z.ai/api/coding/paas/v4";
|
||||
apiKey = "{env:ZAI_API_KEY}";
|
||||
};
|
||||
models = {
|
||||
# Primary worker: 1M ctx flat-rate model.
|
||||
"glm-5.3" = {
|
||||
name = "GLM-5.3";
|
||||
limit = {
|
||||
context = 1000000;
|
||||
output = 128000;
|
||||
};
|
||||
};
|
||||
# High-volume parallel scans: speed-tuned GLM on flat rate.
|
||||
"glm-5-turbo" = {
|
||||
name = "GLM-5 Turbo";
|
||||
limit = {
|
||||
context = 200000;
|
||||
output = 128000;
|
||||
};
|
||||
};
|
||||
};
|
||||
};
|
||||
|
||||
litellm = {
|
||||
npm = "@ai-sdk/openai-compatible";
|
||||
name = "LiteLLM (AZ-Gruppe)";
|
||||
|
||||
Reference in new issue
Block a user