-
-
Notifications
You must be signed in to change notification settings - Fork 1.3k
Issues
is:issue state:open
is:issue state:open
Issue creation is restricted in this repository
Search results
Loop engineering: no cost/token spend kill-switch — an autonomous run can only be bounded by turn count or wall-clock, never by money or tokens
bugSomething isn't workingSomething isn't workingclaudeAuto-trigger Claude analysisAuto-trigger Claude analysisdocumentationImprovements or additions to documentationImprovements or additions to documentationenhancementNew feature or requestNew feature or requestStatus: Open.#3003 In MervinPraison/PraisonAI;Loop engineering: the result-aware stall/oscillation detector is unwired, and the active doom-loop detector is fed degenerate data (constant action, success=True)
bugSomething isn't workingSomething isn't workingclaudeAuto-trigger Claude analysisAuto-trigger Claude analysisenhancementNew feature or requestNew feature or requestStatus: Open.#3001 In MervinPraison/PraisonAI;Goal engineering: no acceptance-criteria completion judge — the autonomous loop terminates on keyword/promise heuristics and ignores
agent.goalclaudeAuto-trigger Claude analysisAuto-trigger Claude analysisenhancementNew feature or requestNew feature or requestquestionFurther information is requestedFurther information is requestedStatus: Open.#2999 In MervinPraison/PraisonAI;Agent harness: context compaction never runs inside the tool-calling loop, and token counting is a char/4 heuristic
bugSomething isn't workingSomething isn't workingclaudeAuto-trigger Claude analysisAuto-trigger Claude analysisenhancementNew feature or requestNew feature or requestStatus: Open.#2991 In MervinPraison/PraisonAI;Autonomous self-improvement (skill review, auto-memory, auto-learning) runs inline on the reply path — add post-response background execution
claudeAuto-trigger Claude analysisAuto-trigger Claude analysisenhancementNew feature or requestNew feature or requestStatus: Open.#2985 In MervinPraison/PraisonAI;Bundled gateway client ignores the negotiated Hello handshake (still joins via legacy frame, never reads advertised features/policy)
bugSomething isn't workingSomething isn't workingclaudeAuto-trigger Claude analysisAuto-trigger Claude analysisenhancementNew feature or requestNew feature or requestStatus: Open.#2983 In MervinPraison/PraisonAI;Knowledge() silently degrades mem0/MongoDB vector-store configs to keyword-only SQLite search due to a swallowed adapter-construction TypeError
bugSomething isn't workingSomething isn't workingclaudeAuto-trigger Claude analysisAuto-trigger Claude analysisStatus: Open.#2972 In MervinPraison/PraisonAI;Memory.store_short_term_async / store_long_term_async crash with TypeError due to a compute_quality_score() signature clash
bugSomething isn't workingSomething isn't workingclaudeAuto-trigger Claude analysisAuto-trigger Claude analysisStatus: Open.#2971 In MervinPraison/PraisonAI;Duplicate: eval/media.py reimplements grader score parser (and drops the 1-10 clamp)
claudeAuto-trigger Claude analysisAuto-trigger Claude analysisdocumentationImprovements or additions to documentationImprovements or additions to documentationStatus: Open.#2970 In MervinPraison/PraisonAI;Duplicate: sync/async after-agent hook share ~40 byte-identical lines
claudeAuto-trigger Claude analysisAuto-trigger Claude analysisStatus: Open.#2969 In MervinPraison/PraisonAI;Duplicate: estimate_tokens() reimplemented four times inside rag/
claudeAuto-trigger Claude analysisAuto-trigger Claude analysisStatus: Open.#2968 In MervinPraison/PraisonAI;Hot-path: agent-lifecycle hook inputs built unconditionally on every chat()/achat()
claudeAuto-trigger Claude analysisAuto-trigger Claude analysisStatus: Open.#2967 In MervinPraison/PraisonAI;