Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.
Python agentaianthropicclaude-codecompressioncontext-engineeringcontext-windowcursorfastapilangchainllmmcpopenaiprompt-engineeringproxypythonragtoken-optimizationtokenstypescript
Fix mypy no-any-return errors in kompress_compressor.py
chopratejas18e31be23a64
diff --git a/headroom/transforms/kompress_compressor.py b/headroom/transforms/kompress_compressor.pyindex aef7796fe..6918d6e0b 100644--- a/headroom/transforms/kompress_compressor.py+++ b/headroom/transforms/kompress_compressor.py@@ -81,7 +81,7 @@ class HeadroomCompressorModel(nn.Module):borderline = (token_probs > 0.3) & (token_probs <= 0.5)keep = token_keep | (borderline & span_boost)- return keep+ return keep # type: ignore[no-any-return]def get_scores(self, input_ids: torch.Tensor, attention_mask: torch.Tensor) -> torch.Tensor:"""Get per-token importance scores (for ranking when target_ratio is set)."""@@ -89,7 +89,7 @@ class HeadroomCompressorModel(nn.Module):hidden = self.encoder(input_ids, attention_mask=attention_mask).last_hidden_statetoken_probs = torch.softmax(self.token_head(hidden), dim=-1)[:, :, 1]span_scores = self.span_conv(hidden.transpose(1, 2)).squeeze(1)- return token_probs * (0.5 + 0.5 * span_scores)+ return token_probs * (0.5 + 0.5 * span_scores) # type: ignore[no-any-return]# ── Model Loading ─────────────────────────────────────────────────────