chat.md says forgelm chat accepts a GGUF file via llama.cpp; no such loading path exists (CR-A13-006) #449

Wave 1 · theme D1-8: Docs describing features that do not exist (doc side of code findings) · highest severity High · 1 finding(s) · effort trivial

Summary

An operator follows the manual to chat with the exported GGUF and gets a traceback on the training-error exit code.

Primary finding

CR-A13-006 — chat.md says forgelm chat accepts a GGUF file via llama.cpp; no such loading path exists

severity High · reachability default · intent contract-violation · confidence confirmed · independently confirmed · reproduced end to end

What happens. run_chat calls inference.load_model, which only uses transformers/unsloth from_pretrained with no gguf_file or llama.cpp backend (inference.py:41-117); a .gguf path fails with a config-JSON error and exit 2, contrary to chat.md:56 (EN/TR).

Where.

• docs/usermanuals/en/deployment/chat.md:56 (confirmed)
• docs/usermanuals/tr/deployment/chat.md:56 (the TR GGUF bullet is at line 56 of the quoted range, shown around line 55-56)
• forgelm/inference.py:83
• forgelm/chat.py:358-366
• forgelm/inference.py:41-117 (tokenizer load at :83, model load at :106)
• forgelm/cli/_parser.py:59-65
• forgelm/cli/subcommands/_chat.py:20-34
• forgelm/inference.py:105

Evidence.

• forgelm chat t.gguf < /dev/null -> raw traceback ending 'OSError: It looks like the config file at 't.gguf' is not a valid JSON file.', exit 2.
• fpy -m forgelm chat ./model.gguf -> exit 2, 'Chat session failed: It looks like the config file at './model.gguf' is not a valid JSON file.'

Recommended fix. Remove the GGUF bullet (point to Ollama or llama.cpp instead), or detect a .gguf suffix and fail with an actionable config error (exit 1).

Acceptance criteria

• A regression test reproduces the failure scenario and asserts the corrected behaviour.
• Documentation listed above is corrected in English and in its Turkish mirror in the same PR.
• Every finding in this issue is resolved or explicitly closed with a reason.

_{Found by the October 2026 ForgeLM code and documentation review against commit cab2947. Code links point at that commit; line numbers may have moved since. Finding IDs (CR-…, DOC-…, TECH-…) are stable references for this backlog.}