Add native Z.ai (GLM) provider

[jiunshinn]

Summary

• Add Z.ai (Zhipu) as a native LLM provider, enabling direct use of GLM models (glm-4.7, glm-5, etc.) without intermediaries like OpenRouter
• Z.ai's API is OpenAI-compatible, so implementation reuses existing message conversion and response parsing
• Fix fallbacks resolution to replace defaults when explicitly set in config (previously fallbacks = {} would not clear Anthropic defaults)

Changes

• src/config.rs: Add zhipu_key to config structs, env/TOML resolution, onboarding wizard; fix fallbacks field to use Option<HashMap> so explicit empty config clears defaults
• src/llm/manager.rs: Add "zhipu" match arm in get_api_key()
• src/llm/model.rs: Add call_zhipu() method and route "zhipu" provider in attempt_completion()

Usage

[llm]
zhipu_key = "env:ZHIPU_API_KEY"

[defaults.routing]
channel = "zhipu/glm-4.7"
worker = "zhipu/glm-4.7"

Test plan

• cargo check passes
• cargo test --lib — all 41 tests pass
• Manual: configure ZHIPU_API_KEY + zhipu/glm-4.7 routing, verify completions work via Discord

[gempir]

How do you use this?

I have the
GLM Coding Lite-Monthly Plan

which worked fine on OpenClaw, but for spacebot i'm getting

errorCompletionError: ProviderError: zhipu/glm-4-plus failed after 3 attempts: ProviderError: Z.ai API error (429 Too Many Requests): Insufficient balance or no resource package. Please recharge.
modelzhipu/glm-4-plus

Cannot change the model to anything but zhipu/glm-4-plus in the Agent Model Routing settings