z.ai models almost always return empty result

[irfansofyana]

Hey there!

not sure if this is really LLM provider issue or summarize tool issue. But I always got "LLM returned an empty summary"

trying the verbose logs, it returns

summarize https://docs.z.ai/devpack/tool/claude#manual-configuration --json --verbose
[summarize] config url=https://docs.z.ai/devpack/tool/claude#manual-configuration timeoutMs=120000 youtube=auto firecrawl=auto length=xl maxOutputTokens=none retries=1 json=true extract=false format=text preprocess=auto markdownMode=off model=zai/glm-4.7 videoMode=auto stream=off plain=false
[summarize] configFile path=/home/irfansofyana/.summarize/config.json model=zai/glm-4.7
[summarize] env xaiKey=false openaiKey=false zaiKey=true googleKey=false anthropicKey=false openrouterKey=false apifyToken=false firecrawlKey=false
[summarize] markdown htmlRequested=false transcriptRequested=false provider=none
[summarize] extract start
[summarize] cache hit extract
[summarize] extract done strategy=html siteName=Overview - Z.AI DEVELOPER DOCUMENT title=Claude Code - Overview - Z.AI DEVELOPER DOCUMENT transcriptSource=none
[summarize] extract stats characters=2898 words=436 transcriptCharacters=none transcriptLines=none
[summarize] extract firecrawl attempted=false used=false notes=none
[summarize] extract markdown requested=false used=false provider=none notes=none
[summarize] extract transcript textProvided=false provider=none attemptedProviders=none notes=none
[summarize] cache miss transcript
[summarize] cache miss summary
[summarize] LLM empty output for zai/glm-4.7; retry 1/1 in 690ms.

Error: LLM returned an empty summary (model zai/glm-4.7).
    at generateTextWithModelId (file:///home/irfansofyana/.nvm/versions/node/v24.12.0/lib/node_modules/@steipete/summarize/dist/esm/llm/generate-text.js:259:27)
    at process.processTicksAndRejections (node:internal/process/task_queues:103:5)
    at async summarizeWithModelId (file:///home/irfansofyana/.nvm/versions/node/v24.12.0/lib/node_modules/@steipete/summarize/dist/esm/run/summary-llm.js:24:20)
    at async Object.runSummaryAttempt (file:///home/irfansofyana/.nvm/versions/node/v24.12.0/lib/node_modules/@steipete/summarize/dist/esm/run/summary-engine.js:151:28)
    at async runModelAttempts (file:///home/irfansofyana/.nvm/versions/node/v24.12.0/lib/node_modules/@steipete/summarize/dist/esm/run/model-attempts.js:23:22)
    at async summarizeExtractedUrl (file:///home/irfansofyana/.nvm/versions/node/v24.12.0/lib/node_modules/@steipete/summarize/dist/esm/run/flows/url/summary.js:280:32)
    at async runUrlFlow (file:///home/irfansofyana/.nvm/versions/node/v24.12.0/lib/node_modules/@steipete/summarize/dist/esm/run/flows/url/flow.js:297:9)
    at async runCli (file:///home/irfansofyana/.nvm/versions/node/v24.12.0/lib/node_modules/@steipete/summarize/dist/esm/run/runner.js:509:9)
    at async runCliMain (file:///home/irfansofyana/.nvm/versions/node/v24.12.0/lib/node_modules/@steipete/summarize/dist/esm/cli-main.js:105:9)

[omniwired]

@steipete except 4.5-Flash

The issue is specific Z.AI models fail. Either:

1. Your Z.AI API key doesn't have access to those models. Mine is the coding plan
2. Summarize's model config is wrong for those models (they're set to use anthropic-messages API instead of OpenAI)

Working model: zai/glm-4.5-flash

[vlnd0]

hey, lifehack:
if you have z.ai subsciption plan - just mimic under the claude code

.summarize/config

{
  "model": "anthropic/claude-sonnet-4-5",
  "anthropic": {
    "baseUrl": "https://api.z.ai/api/anthropic"
  }....

and ANTHROPIC_API_KEY with z.ai key

[smeeklai]

Having the same issue