修改agent流程，分步构建cdsl

backend/README.md

@@ -10,39 +10,63 @@ The backend owns the application API and CAD generation workflow:
 
 Expected development entrypoint: `app.main:app`, served by Uvicorn.
 
-## Incremental Generation Configuration
+## Reasoning Effort
 
-Incremental generation is enabled by default. It requires a separately
-configured vision-capable review model and the Python OpenCascade/Pillow
-technical renderer; a run
-fails instead of skipping visual review when either is unavailable.
+The backend uses the Chat Completions API. Configure a provider's reasoning
+budget with `CDSL_<PROVIDER>_REASONING_EFFORT`; for the current OpenAI setup:
 
 ```dotenv
-# Authoring provider/model must already be configured as usual.
-CDSL_INCREMENTAL_GENERATION=1
+CDSL_OPENAI_REASONING_EFFORT=medium
+```
 
+Use `low`, `medium`, or `high` according to the latency/cost versus quality
+tradeoff. The setting is sent as Chat Completions' `reasoning_effort` field to
+authoring, streaming, and visual-review requests. Leave it empty to use the
+provider/model default. The selected OpenAI-compatible endpoint must support
+the requested value.
+
+## Autonomous CDSL Agent Configuration
+
+The autonomous agent writes one frozen free-form `requirements.md`, then
+observes, measures, renders and appends one CDSL feature at a time. Its author
+uses normal function calls; no provider strict JSON Schema capability or
+complete modelling DAG is required. Candidate fragments are rebuilt in a
+staging directory through `cdsl_only` before a checkpoint can be committed.
+
+Final publication requires a separately configured vision-capable review model
+and the Python OpenCascade/Pillow technical renderer. The agent may build and
+inspect intermediate checkpoints without image review; a final run fails
+closed if its independent review configuration is unavailable.
+
+```dotenv
 # Must name one configured provider and one model listed in that provider's
 # CDSL_<PROVIDER>_VISION_MODELS setting. It is intentionally not inferred
 # from the authoring model.
-CDSL_REVIEW_PROVIDER=openai
-CDSL_REVIEW_MODEL=gpt-4.1-mini
-CDSL_OPENAI_VISION_MODELS=gpt-4.1-mini
+CDSL_REVIEW_PROVIDER=deepseek
+CDSL_REVIEW_MODEL=deepseek-v4-flash-vision-exp
+CDSL_DEEPSEEK_VISION_MODELS=deepseek-v4-flash-vision-exp
 
 # Install Python rendering dependencies. The renderer reads the revision STEP
 # file and creates canonical images without a browser or GPU driver.
 pip install -r requirements.txt
 
-# Optional per-node retry budgets.
-CDSL_NODE_AUTHORING_ATTEMPTS=2
-CDSL_NODE_REPAIR_ATTEMPTS=2
-CDSL_NODE_REPLAN_ATTEMPTS=1
+# Limits apply to the current checkpoint head, never to total task complexity.
+CDSL_AGENT_TOOL_CALLS_PER_CYCLE=12
+CDSL_AGENT_CANDIDATE_ATTEMPTS_PER_HEAD=3
+CDSL_AGENT_CONSECUTIVE_NO_PROGRESS_LIMIT=6
+CDSL_AGENT_MAX_FEATURES_PER_FRAGMENT=6
+CDSL_AGENT_CONTEXT_CHAR_LIMIT=24000
+CDSL_AGENT_RENDER_CACHE=true
 ```
 
-Every checkpoint is rebuilt from its fully materialized CDSL through the
+The author chooses each coherent 1-6 feature batch. Every rebuilt batch is
+rendered and independently reviewed before it can become a checkpoint; only
+an accepted reviewer verdict advances the working model. Every checkpoint is rebuilt from its fully materialized CDSL through the
 `cdsl_only` runtime. Checkpoint GLB files are preview-only; STEP, CDSL, and
 reports are available only after the task reaches `COMPLETED`.
 
-The generation plan contains semantic node IDs only. The backend derives the
-unique CDSL feature and sketch IDs from each node, then writes them during
-fragment materialization. This keeps naming and topology ownership stable
-without requiring the authoring model to reproduce internal identifiers.
+The backend assigns feature and sketch IDs, appends causal dependencies and
+expands only opaque current-snapshot selector tokens. It does not compile
+geometry templates or correct workplanes, profiles, sizes, directions or
+boolean semantics authored by the model. Failed candidates remain auditable
+but never become revisions.