diff options
| author | Paul Buetow <paul@buetow.org> | 2025-09-03 16:08:49 +0300 |
|---|---|---|
| committer | Paul Buetow <paul@buetow.org> | 2025-09-03 16:08:49 +0300 |
| commit | 7088db8f1c77a08ad37a1bd0437a0c066dae80fa (patch) | |
| tree | ff4454a0acbf010791999599a91674d3170b3742 /docs/configuration.md | |
| parent | 7e7908f01a47cf92b3e9b093083e2b9cc1b4c7db (diff) | |
Docs: add LSP completion tuning section and manual_invoke_min_prefix\n\n- Expand docs/configuration.md with guidance for debounce/throttle\n- Document manual_invoke_min_prefix and related env vars
Diffstat (limited to 'docs/configuration.md')
| -rw-r--r-- | docs/configuration.md | 9 |
1 files changed, 9 insertions, 0 deletions
diff --git a/docs/configuration.md b/docs/configuration.md index 72ac92e..e5e7dfa 100644 --- a/docs/configuration.md +++ b/docs/configuration.md @@ -44,6 +44,7 @@ Key fields: - log_preview_limit: max characters of context preview logged. - completion_debounce_ms: minimum idle time before sending completion requests. - completion_throttle_ms: minimum spacing between completion requests (0 disables). +- manual_invoke_min_prefix: minimum typed identifier chars required for manual invoke to proceed without structural triggers (0 allows always). - no_disk_io: avoid reading files from disk when building context. - trigger_characters: LSP completion trigger characters. - coding_temperature: optional override for LSP calls. @@ -101,6 +102,14 @@ Notes: - Alternatively, run Ollama in OpenAI‑compatible mode and use the OpenAI provider with `openai_base_url` pointed at your local endpoint. +## LSP completion tuning + +- Debounce: `completion_debounce_ms` waits until there has been no recent input for at least this many milliseconds before sending a completion request. Recommended 150–300 ms to balance responsiveness and API usage. +- Throttle: `completion_throttle_ms` enforces a minimum spacing between completion requests, across both chat and provider-native paths. Set to 0 to disable. Recommended 300–600 ms if you still see excessive requests with just debounce. +- Manual invoke prefix: `manual_invoke_min_prefix` requires this many identifier characters before a manual completion (TriggerKind=1) proceeds without other triggers. Use 0 to always allow manual invoke. + +Environment variables mirror these settings: `HEXAI_COMPLETION_DEBOUNCE_MS`, `HEXAI_COMPLETION_THROTTLE_MS`, `HEXAI_MANUAL_INVOKE_MIN_PREFIX`. + ## Temperature behavior - What it is: controls randomness/creativity of outputs. |