marfrit/aish
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix: buffer exec output for next user turn; alternation for strict templates

Browse Source 
User-test surfaced the bug: with `deep` (mistral-nemo-12b) active,
running `list files` -> y on `CMD: ls` -> `Are there directory entries
beginning with "lor"?` returned a Jinja exception:

    api: ... Error: Jinja Exception: After the optional system message,
    conversation roles must alternate user/assistant/user/assistant/...

Cause: §6 specified "exec output injected into context uses role 'user'
with a prefix tag '[exec output]'." This works for permissive templates
(qwen2.5-coder-1.5b, the `fast` preset) but produces a back-to-back
user/user pair on strict templates that enforce the OpenAI alternation
contract — `[exec output]` user turn followed by the user's actual
follow-up question.

Fix:

context.lua:
  - new field `pending_exec_output` (initially nil)
  - new method `:append_exec_output(out)` buffers (concat on subsequent
    captures so multi-shell-then-ai still merges everything)
  - new method `:append_user(content)` flushes buffered exec output as
    a `[exec output]\n...\n\n` prefix and appends a user turn
  - `:reset()` also clears the buffer

repl.lua:
  - run_shell calls ctx:append_exec_output(out) instead of
    ctx:append({role="user", content="[exec output]\n"..out})
  - ask_ai calls ctx:append_user(text) instead of raw :append; saves
    prev_pending so a broker error can restore the buffer for retry

PHASE0.md §6:
  - amended the role-injection paragraph to describe the buffer-and-
    prepend policy; the §3 invariants list is untouched (this was a §6
    design detail, not a locked invariant)

Verification:
  - context unit tests cover: alternation after the failing sequence,
    multi-shell merge, reset clears buffer, broker-error retry path
  - live reproduction against `deep` (mistral-nemo) of the exact
    user-reported sequence succeeds; model responds with a sensible
    `CMD: ls | grep '^lor'` instead of a Jinja exception

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
Markus Fritsche
2026-05-10 18:41:21 +00:00
parent 8870eb0451
commit 16490e6905
3 changed files with 42 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/PHASE0.md
+3 -1
View File
@@ -141,7 +141,9 @@ Each turn is stored in `context.lua` as:
{ role = "system" | "user" | "assistant", content = "..." }
```
The system prompt is prepended on every request and is not stored as a history turn. Exec output injected into context uses role `"user"` with a prefix tag `[exec output]`.
The system prompt is prepended on every request and is not stored as a history turn.
**Exec output injection.** Captured shell-exec output is **not** appended as its own user turn — that produces user/user back-to-back, which strict chat templates (e.g. `mistral-nemo-instruct`'s Jinja) reject with `roles must alternate user/assistant/...`. Instead, exec output is buffered on the context and prepended to the **next** user turn with a `[exec output]` tag. Multiple shell calls between AI turns concatenate. `:reset` clears the buffer. The user-visible behavior is unchanged; only the role alternation seen by the broker differs.
### System prompt (Phase 0 default)