Skip to content

Commit 23ffcf6

Browse files
cathleenyIsaac
andcommitted
Merge main into fix/kernel-honor-https-port
Resolve the CHANGELOG.md conflict by keeping both Unreleased entries: this PR's kernel port entry and main's Arrow row materialization entry. No code conflicts. Co-authored-by: Isaac <no-reply@databricks.com> Signed-off-by: Cathleen Yan <58714163+cathleeny@users.noreply.github.com>
2 parents 468af3a + 1c499e2 commit 23ffcf6

15 files changed

Lines changed: 1101 additions & 105 deletions

File tree

‎.bot/config.yaml‎

Lines changed: 10 additions & 32 deletions
Original file line numberDiff line numberDiff line change
@@ -1,8 +1,4 @@
1-
# Engineer-bot task config for databricks-sql-nodejs (the Databricks SQL connector
2-
# for Node.js/TypeScript). Loaded by
3-
# databricks_bot_engine.engineer_bot.bot_config.load_bot. The reviewer needs NO
4-
# .bot/config.yaml (its prompts are engine-owned + the optional additive
5-
# .bot/prompts/review/system.md); only the engineer-bot is config-driven.
1+
# Engineer-bot task config and shared retrospective lesson routing.
62
name: databricks_sql_nodejs_bugfix
73
marker_namespace: engineer-bot-nodejs
84
bot_login_prefix: peco-engineer-bot
@@ -97,11 +93,6 @@ author:
9793
issue_url: ISSUE_URL
9894
context_files:
9995
- issue_body.txt # {{issue_body}}
100-
# Close the loop: the author phase reads this log back so fixes benefit from
101-
# what the retrospective has learned. MUST match `retrospective.log_path` below
102-
# — the retrospective WRITES that path; the author only READS it if the two
103-
# agree. (Set together, they connect learning → application.)
104-
knowledge_log: .claude/knowledge/learning-log.md
10596

10697
# Engine orchestration for the author phase. `bug-fix` runs the plan →
10798
# author_tests → fix pipeline (write a failing test → fix the code → re-run to
@@ -111,27 +102,14 @@ author:
111102
# prompts/engineer/system.md, not here.
112103
flow: bug-fix
113104

114-
# Daily learning extraction (retrospective flow). Run by engineer-bot-learning.yml
115-
# via `python -m databricks_bot_engine.engineer_bot.retrospective`: over an adaptive
116-
# look-back window it sweeps every merged PR (diff + review comments, which the
117-
# engine gathers ITSELF per PR) plus engineer-bot author runs, and if the model
118-
# finds a durable, reusable learning, opens/updates a single rolling PR appending a
119-
# dated section to log_path. Human-gated — never commits the canonical log directly.
120-
# Omitting this block makes the retrospective a no-op.
121-
#
122-
# `system_prompt` is intentionally OMITTED: this repo ships no
123-
# prompts/retrospective_system.md, so the engine's built-in base prompt is used.
124-
# (A set-but-missing system_prompt file would be a hard error; an UNSET key uses
125-
# the base. Add the key later only if you write a repo-specific additive prompt.)
126-
# The log is seeded at `log_path` (see `.claude/knowledge/learning-log.md`) so the
127-
# author phase has something to read before the first retrospective PR lands; the
128-
# retrospective appends dated sections to that same file thereafter.
105+
# Both bots receive matching lessons; the old log is a frozen archive.
129106
retrospective:
130-
# No `model:` key: it is RETIRED in the engine (retrospective.py warns
131-
# "`retrospective.model` is no longer honored") — the learning run always uses
132-
# the engine-owned engineer-bot model (databricks-claude-opus-4-8[1m], override
133-
# via the ENGINEER_BOT_MODEL_OVERRIDE workflow env). Setting it here only emits a
134-
# per-run ::warning::. Matches the other driver repos' retrospective blocks.
135-
log_path: .claude/knowledge/learning-log.md # MUST match author.knowledge_log above
136-
branch_prefix: ai/learning-pr-
107+
log_mode: distill
108+
log_path: .claude/knowledge/learning-log.md
109+
journal_dir: .claude/knowledge/journal
110+
lessons_path: .claude/knowledge/lessons.md
111+
auto_merge_learning_pr: false
112+
author_workflow: engineer-bot.yml
113+
# Keep the legacy append-mode PR (#500) on its original branch.
114+
branch_prefix: ai/learning-distill-pr-
137115
pr_label: engineer-bot-learning
Lines changed: 10 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,10 @@
1+
# Learning journal 2026-11-18
2+
3+
### 2026-11-18: Bot workflow and CI — GitHub Actions dot dereference accepts hyphenated property names
4+
**Context:** In PR #620 the seeded lesson L-0060 said hyphenated workflow_dispatch inputs need bracket notation. In PR #621 it was corrected: GitHub Actions expressions allow `-` in property names under dot dereference, so the repo's own `${{ inputs.window-hours }}` is valid. The archived #499 claim that it parses as subtraction was wrong.
5+
**Rule:** Do not flag or rewrite `${{ inputs.some-name }}` as a subtraction bug, because GitHub Actions expressions allow hyphens in dot-dereferenced property names.
6+
7+
### 2026-11-18: Retrospective / lessons — Check a lesson against nearby code and docs before promoting it to a standing rule
8+
**Context:** In PR #620, archived log entries were distilled into lessons.md, which is now the only learning text injected into both bots' prompts. A reviewer found that L-0060 contradicted the workflow file right next to it and was unsupported, and PR #621 cut it back to the verified part only.
9+
**Rule:** Before promoting an archived or journal observation into lessons.md, check it against the current code and the official docs, and keep only the verified part, because injected lessons drive bot edits and findings.
10+

0 commit comments

Comments
 (0)